Animated Presentation of Static Infographics with InfoMotion

Eurographics Conference on Visualization (EuroVis) 2021

R. Borgo, G. E. Marai, and T. von Landesberger

(Guest Editors)

Volume 40 (2021), Number 3

Yun Wang

, Yi Gao

1,2

, Ray Huang

, Weiwei Cui

, Haidong Zhang

, and Dongmei Zhang

Microsoft Research Asia

Nanjing University

Figure 1:

(a) An example infographic design. (b) The animated presentations for this infographic show the start time, duration, and animation

effects applied to the visual elements. InfoMotion automatically generates animated presentations of static infographics.

Abstract

By displaying visual elements logically in temporal order, animated infographics can help readers better understand layers of

information expressed in an infographic. While many techniques and tools target the quick generation of static infographics, few

support animation designs. We propose InfoMotion that automatically generates animated presentations of static infographics.

We ﬁrst conduct a survey to explore the design space of animated infographics. Based on this survey, InfoMotion extracts

graphical properties of an infographic to analyze the underlying information structures; then, animation effects are applied to

the visual elements in the infographic in temporal order to present the infographic. The generated animations can be used in

data videos or presentations. We demonstrate the utility of InfoMotion with two example applications, including mixed-initiative

animation authoring and animation recommendation. To further understand the quality of the generated animations, we conduct

a user study to gather subjective feedback on the animations generated by InfoMotion.

CCS Concepts

• Human-centered computing → Visualization design and evaluation methods;

1. Introduction

Combining diverse visual elements, infographics have become a per-

vasive means of improving the comprehension of complex informa-

tion. Enhanced by well-crafted animations, infographics can present

contents progressively and narratively to facilitate perception of

information [Fis10]. In addition, creative animations can guide view-

ers’ attention and improve user engagement [CRP

∗

16]. Professional

designers and artists create animated infographics to expand the

inﬂuence of content. Non-experts such as marketers and government

agencies also leverage animations to create informative, aesthetic,

compelling, and impressive presentations. However, the design of

animated infographics requires tremendous effort [ARL

∗

17]. Creat-

ing an elaborate animated infographic involves multiple steps. One

needs to carefully plan the timings and effects of animations.

Consider a hypothetical example in which a nutritionist, Diana, is

preparing a presentation to illustrate Mediterranean diet (Figure 1a).

To leave a great impression to the audience, she goes through design

examples of animated infographics, and imagines an animation like

this: First, the central icon and title show. Then, every slice of the

donut spin clockwise one by one. When the slice is spinning in,

the links wipe in at the same time. For each slice, when the links

are completely shown, the icons, numbers, and descriptions should

gradually appear one by one. To implement the animation design,

she needs to carefully draw out a timeline to arrange the start time

points for each animation (Figure 1b), and apply animation effects

Wiley & Sons Ltd. Published by John Wiley & Sons Ltd.

Yun Wang et al. / Animated Presentation of Static Infographics with InfoMotion

to each visual element. If the resulting animation is not satisfactory

enough, she needs to repeat this process iteratively through trial and

error. The entire process can be tedious and laborious. Non-expert

users without an animation design background may not have clear

animation designs in mind and be hindered by such complicated

settings. They may compromise by simply using a monotonous

animation design, e.g., adopting the same effect and showing all the

elements at once, though they might intend to create more elaborated

ones with abundant animation effects and carefully designed time

arrangements to display and convey their messages.

Visual elements in an infographic compose information structures

implicitly. For example, in Figure 1, an assembly of slice, percent-

age, icon, etc., composes one repeating unit. With similar visual

representations, each unit presents a single part of the whole story,

and at the same time, they are combined to tell the whole story. By

manually coding of 965 infographics, researchers [LWL

∗

20] ﬁnd

the majority (64.1%) of infographics contain clear narrative ﬂows.

People inject visual hints to imply narrative ﬂows and relations of

visual elements. One step further, we aim to infer semantic roles

of each visual element in these static infographics and organizing

them into information structures, so that we can arrange animation

sequences and effects accordingly.

In this paper, we propose an automatic approach of animation

design, InfoMotion, that recognizes underlying structures of static

infographics and further applies animation sequences and effects,

based on the underlying information structures, to enhance presen-

tations of infographics. First, we conduct a survey on real-world

designs of animated infographics and summarize the design of an-

imation into design primitives to compose information structures.

Then, we propose InfoMotion, a technique that generates animated

presentations of static infographics. InfoMotion infers information

structure from a static infographic by analyzing visual properties of

the elements; presentation orders and animation effects are arranged

to animate the visual elements. We report quantitative evaluations

of our structure inference on 120 infographic examples. We demon-

strate the utility of InfoMotion with two example applications, i.e.,

mixed-initiative animation authoring and animation recommenda-

tion. We further conduct a user study to gather subjective feedback

on the animations generated by InfoMotion and understand the dif-

ferences between auto-generated animations and animations crafted

by designers. Participants shown no preference on the 12/18 anima-

tion pairs generated by InfoMotion and crafted by designers.

2. Related Work

2.1. Infographics

Infographics have gained increasing interest among researchers

in the ﬁeld of visualization. Combining data content and visual

embellishments, infographics can engage and impress users eas-

ily [BMG

∗

10, HKF15, BVB

∗

13, WSK

∗

19]. Researchers have ex-

plored automated approaches to understand infographics from the

perspective of visual importance [BKO

∗

17], personality [ZCL18],

and quality [FWD

∗

19,WXL

∗

21]. Recently, Lu et al. [LWL

∗

20] ana-

lyze the presentation ﬂow of static infographics leveraging indexes

(numbers and letters) extracted from bitmap images. Inspired by this

work, we further propose an algorithm to reconstruct information

structures so as to enable animation arrangement for static info-

graphics. On a related note, Thompson et al. recently summarize the

design space of animated visualizations [TLLS20]. However, the

study only focuses on summarizing low-level primitives of anima-

tion design. There is a lack of empirical studies to understand how

animation creators present infographics gradually to reveal informa-

tion. To this end, we collect a dataset of animated infographics and

analyze the design space of presenting animated infographics.

Authoring tools and automated generation techniques have

been proposed to assist in creating more expressive visualizations

[WWS

∗

21]. For example, Data Illustrator [LTW

∗

18] and Charticula-

tor [RLB19] help with the design of bespoke visualizations through

data binding applied to vector graphic properties. Researchers have

also proposed design tools for expressive infographics with im-

ages, icons, and hand-drawn shapes. For example, DDG [KSL

∗

17],

DataInk [XHRC

∗

18], and InfoNice [WZH

∗

18] enable data bind-

ing with custom shapes, hand-drawn shapes, and icons, images,

and texts, respectively. Automated approaches are also explored to

generate infographic data stories, e.g., organizing data facts into

fact sheets [WSZ

∗

20], generating infographics from natural lan-

guage statements [CZW

∗

20], or extracting timeline templates from

infographic images [CWW

∗

19]. While the authoring of static info-

graphics is well-studied, there is no existing technique or tool that

supports the animating of expressive visualizations or infographics

with rich design elements such as icons and embellishments. In our

work, we propose InfoMotion to generate animated presentation of

static infographics.

2.2. Animation Design in Visualization

Animations are commonplace in visualization and have been re-

searched for a long time. Early work by Tversky et a. [TMB02]

ﬁnds although animations might not be effective for data analysis,

they are attractive in presentations. Animations can serve a wide vari-

ety of purposes in presentations, such as revealing data relationships,

helping with orientation, catching attention, etc [CRP

∗

16, BS90].

In the context of visual analytics, animated transition techniques

have been adopted to keep users oriented during changes and manip-

ulations of complex datasets, such as dynamic networks [BPF14],

streaming data [HVF13], multidimensional data [EDF08], and doc-

ument histories [CDBF10]. To improve the tracking of moving

objects, researchers have explored animation strategies such as stag-

ing [CRP

∗

20], trajectory bundling [DCZL15], and temporal distor-

tion [DBJ

∗

11]. Described as visually pleasing and engaging, anima-

tion is also an effective method of vividly telling stories to improve

understandability and enhance user experience [HR07,ARL

∗

15] .

For example, Ruchikachorn et al. [RM15] propose teaching visu-

alizations by linking a new one to another more familiar one and

change through animated morphing; Kim et al. [KCH19] design

animated visualizations to convey common aggregation operations;

Wang et al. [WCL

∗

16] design animated narrative visualization to

present video clickstream data.

More recently, Ge et al. [GZL

∗

20] and Kim et al. [KH20] propose

high-level languages that enable users to specify the animations

for data charts. Although the proposed techniques also generate

animated graphics for static ones, they focus on data charts that are

represented by well-structured forms (data-enriched SVG or Vega-

Yun Wang et al. / Animated Presentation of Static Infographics with InfoMotion

lite), where the roles (e.g., marks, axes) of each visual element (e.g.,

lines, circles) are clearly assigned. However, the visual elements in

an infographic are usually not structured semantically. Likewise, we

propose information structures that animations can be organized

and arranged accordingly.

InfoMotion takes a reverse-engineering way to construct infor-

mation structures. While existing work on reverse-engineering vi-

sualizations constructs visualization structures by analyzing vector

graphics [SKC

∗

11, SHL

∗

16, PH17, JKS

∗

17, WTD

∗

20], they only

target standard chart templates such as bar charts and line charts.

InfoMotion more ﬂexibly identiﬁes repeating units and construct

information structures to cover a variety of infographics. Through

this way, different animation strategies can be further conﬁgured to

cater various needs of expressive animated presentations.

Authoring tools have sought to empower people to create ani-

mated infographics. Commercial tools such as Flourish [ﬂo20] and

Visme [Vis20] support animations through a template-editing man-

ner, where system developers need to create animated infographic

templates in advance. DataClips [ARL

∗

17] similarly provides a set

of templates for users to craft data-driven video. However, with

the template-editing approach, the number of templates limits the

richness of resulting animations. The animations created with a

particular tool may look similar. Presentation tools like PowerPoint,

provide more ﬂexible templates (e.g., SmartArt) with information

structure embedded. Users can conﬁgure animations by choosing

effect options (i.e., all at once, one by one, and as one object). How-

ever, it is still a closed system where users cannot add new templates.

When InfoMotion is leveraged by the developers of template-editing

tools, animation templates can be created automatically from static

infographics; when InfoMotion is integrated into authoring tools, it

can empower the tools to generate animation design candidates and

help users create animated presentation easily.

3. Survey of Animated Infographic Presentations

Infographics are a composition of visual elements. The variety of

visual elements is rich. Within an infographic design, the author

may use different icons, colors, and texts to convey information.

To ease the interpretation of data and information, infographic au-

thors may inject visual hints to guide readers to trace the reading

ﬂows [DMM04]. Lu et al. [LWL

∗

20] found that 64.1% infograph-

ics of a large infographic dataset with 965 infographics contain

clear narrative ﬂow, where visual elements are organized into higher

level visual units to convey messages. Because of the complexity of

"other" infographics, as a ﬁrst step, we scope our research on this

majority type of infographics that enumerate information pieces in

parallel or sequentially.

An effective animated presentation of infographics arranges the

appearance of visual elements in temporal order to guide users’ atten-

tion and convey data and information clearly and logically [CRP

∗

16],

which, however, involves a lot of low-level controls over visual ele-

ments. Our goal with InfoMotion is to facilitate the easy creation of

meaningful animations while striking a balance between ﬂexibility

and expressiveness. Therefore, we ﬁrst conduct a survey to under-

stand common design practices in animated infographics crafted by

human designers that organize visual elements into effective and

meaningful animation sequences.

2016

TITLE 03

2015

TITLE 02

2014

TITLE 01

2017

TITLE 04

History of Architecture

www.historyofarchitecture.nowhere

Figure 2:

An example infographic design. The design includes (a)

title, (b) main body, and (c) footnote. (d) is one of the repeating

units and (e) is one of the connectors. The element groups of this

infographic include (f) unit title group, (g) unit description group,

(h) unit index group, (i) unit icon group, and (j) background group.

3.1. Dataset Collection

We collected animated infographics in different forms, including

data videos, animation source ﬁles such as PowerPoint ﬁles, and

online animation tutorials. We searched through animated info-

graphic examples from video resource sites such as youtube.com

and vimeo.com. We also searched via Google for animated info-

graphics and downloaded presentation template ﬁles in the form of

PowerPoint and Keynote slides. We use keywords including the com-

binations of “infographic”, “animation”, “tutorial”, “presentation”,

“infographic PowerPoint template”, “infographic After Effects tem-

plate”, etc. We manually went through search results and selected

the infographics that ﬁt into the scope of our survey. Overall, we

collected 203 animated infographic designs. Among the designs we

collect, 69.5% are from animation design templates (PowerPoint

and Keynote slides), and 30.5% are from real world videos and GIFs.

We used an open coding strategy and conducted a multi-pass man-

ual coding of the 203 animation designs and iteratively improved

the taxonomy through multiple rounds of discussion among three

authors.

3.2. Information Structure

To understand how to animate static infographics, we need to ana-

lyze static infographic designs and build the information structures.

An infographic design is composed of a set of visual elements, such

as textboxes, shapes, and images, with visual properties, such as

width, height, shape, color, etc. Here, we summarize key attributes

to describe the

information structure

of an infographic design with

an example design (Figure 2):

• Repeating Units

: Multiple visual elements that are combined and

repeated to show parallel or sequential structure in an infographic.

The units usually have similar designs and are placed to certain

positions to imply the relations. For example, Figure 2d is one of

the repeating units.

• Element Groups

: An element group contains semantically simi-

lar elements that belong to different repeating units and contribute

Yun Wang et al. / Animated Presentation of Static Infographics with InfoMotion

Figure 3:

The repeating units in infographics are categorized into four layout patterns: (a) Linear layout [con19,tim18, tut17]; (b) Radial

layout [mil18, 6op19, 9op19]; (c) Segmented layout [pro19, rou19,pen19]; (d) Freeform layout [tim19,bal18,map18].

to the design of each repeating unit. The elements in an element

group are usually visually similar. For example, the unit title el-

ement group have all the unit titles, which have the same fonts;

the unit icon element group have all the unit icons, which have

similar color and sizes.

• Unit Layout

: The spatial arrangement of visual elements is usu-

ally carefully chosen. The locations of repeating units in an info-

graphic determines the unit layout of the infographic. Generally,

the unit layout determines the presentation order.

• Connectors

: Some infographic designs may contain speciﬁc or-

der themselves. Animations are then applied in a particular order

based on infographic designs. These infographics have connec-

tors like lines, arrows to connect visual elements and express

their relations. Then the designs can clearly convey the logic

connections and reading orders of the infographics.

• Semantic Tags

: The roles of visual elements in an infographic

design. For example, textboxes are used as titles (Figure 2a),

footnotes (Figure 2c), etc., for an infographic.

3.3. Infographic Unit Layout

Designers place visual units into different layout patterns, forming

information sequences. Figure 3 shows four main styles of info-

graphic unit layouts we found in our dataset:

• Linear Layouts

(86; 42.36%) are mostly seen in infographic

designs. All the infographic units are arranged into a horizontal,

vertical, or diagonal line. Different from segmented layouts, the

inner layout of each unit should be the same for linear layouts.

• Radial Layouts

(49; 24.14%) refers to the layouts of an info-

graphic with visual units placed together forming an arc or a

circle. It is also a common design in infographics.

• Segmented Layouts

(33; 16.26%) refers to layouts when units

are arranged in a zigzag manner. Usually, the direction of the units

is reversed successively. For example, they may be placed on both

sides along a center line. They can also be placed in horizontal,

vertical, or diagonal ways.

• Freeform Layouts

(35; 17.24%) are layouts when the units are

not put in regular patterns. For example, in the ﬁrst infographic

design in Figure 3d, the text descriptions follow the curved arrows.

Other infographics with clear structure may also have random

layouts for the units. For example, when the units are placed on a

map, they may not form clear layouts. Designers may use arrows

or other visual elements to indicate the internal logic structures

of the elements.

3.4. Animated Presentation Design

We further study the arrangement of temporal animation sequences

for each visual element. Overall, we observe most animations follow

common reading order, e.g, starting from the up-left corner to the

bottom-right corner, based on the layouts. Another major style shows

animations following content semantics. For example, they may

show the title, subtitle, and then the core part.

Animation designers follow the information structure to group

the elements inside the units and apply animations to them as a

whole. Similar to previous research on narrative structures [HKL17],

animation designers develop three main styles for the structured

information (Figure 4). (1)

Concurrently

(20; 9.85%): The ani-

mations are applied to the whole structure of the infographics and

are shown all at once. (2)

By repeating units

(170; 83.74%): The

infographics are divided into similar or repeating units to compose

the infographic structure. This method prioritizes the unit relations

of the elements. Where units are shown one by one and the ele-

ments in a unit tend to appear together. (3)

By element groups

(13; 6.40%): This method prioritizes the element relations across

units. Similar elements are shown together. For example, all the unit

titles may appear at ﬁrst and all the unit descriptions may follow.

Designers sometimes show visual elements by element groups. For

example, designers may show the titles of each unit at ﬁrst and then

the detailed descriptions together.

For those with connectors (97; 47.78%), we categorize the se-

quences into ﬁve types (1)

Connector-ﬁrst

(33; 34.02%): Connec-

tors in the infographic are shown ﬁrst, and then the units, which

is usually applied to connectors passing through all the units; (2)

Content-ﬁrst

(5; 5.15%): Units in the infographic are shown ﬁrst,

and then connectors; (3)

Interleaved

(51; 52.58%): Connectors and

units are shown one by one alternately. (4)

Embellishment-ﬁrst

(2;

2.06%): Embellishments in the infographic are shown ﬁrst, and then

other components; (5)

All at once

(6; 6.19%): Major components

are shown all at once.

Yun Wang et al. / Animated Presentation of Static Infographics with InfoMotion

Unit1

Unit2

Unit3

(b)(a) (c)

Figure 4:

Visual elements are arranged temporally into stages, cor-

responding to the sequences of animations: (a) elements are shown

concurrently; (b) elements are presented by repeating units; (c)

elements are presented by element groups.

Animation pacing, which is usually adopted to arrange animations

into temporal stages and improve understandability and engagement

[TLLS20], is commonly designed for infographics animated by

repeating units and by element groups. Overall, there are three

different styles of animation pacing:

One-by-one

(175; 86.21%):

Animations are merged into groups or units to and shown one by

one. Animation effects can also be different or the same.

All-at-

once

(20; 9.85%): Animations are shown all at once. Designers can

apply one animation effect for all infographic components or apply

different animation effects for different components at the same

time.

Staggering

(8; 3.94%): All animations start at different times

and each animation starts with a constant temporal delay after the

previous one.

4. InfoMotion

Following animated infographics design practices, we propose In-

foMotion, a technique that generates animated infographics auto-

matically from static infographic designs. We describe two stages,

namely, structure inference and presentation arrangement, to gen-

erate animated presentations. At the structure inference stage, Info-

Motion analyzes the visual elements and infers underlying informa-

tion structures from the graphical properties of the visual elements

(width, location, color, shape, etc.). At the presentation arrangement

stage, presentation orders and animation pacing can be arranged

based on design practices summarized from the survey.

The infographic designs are usually in the form of graphic de-

sign ﬁles, in the format of vector image (e.g., svg, pdf, ai, ppt,

etc.), and consist of visual elements, including text boxes, icons,

shapes. InfoMotion takes static vector infographics as input and

parses the infographic designs into a set of visual elements (e.g.,

textboxes, shapes, and icons) and their properties (e.g., position,

color, and size). Although many vector images support the annota-

tion of groups, such as <g> tags in SVG ﬁles, they can be unreliable

and introduce mistakes. Therefore, we treat infographic inputs as a

bag of unstructured visual elements. Taking the properties of these

visual elements, InfoMotion goes through multiple stages to infer

information structures. After that, InfoMotion arranges animation

sequences and applies animation effects to the visual elements. The

animations can be further synthesized by different applications and

rendered to the users. The output of InfoMotion is an abstract speci-

ﬁcation of the animations. Infographic designs are organized into

structures consisting of visual elements. Animations are applied to

each visual element. For each animation, we describe animation

start time and animation effect type. Design tools can directly follow

the animation speciﬁcation to add animations for static infographics.

4.1. Structure Inference

The repeating units that are composed of visual elements usually

have similar designs and are placed in certain positions to imply the

relations. However, it is not easy to identify the elements composing

the units. Visual embellishments with various designs may appear at

any position in an infographic. To tackle these problems, we design

a bottom-up approach to analyze the structure of infographics. Our

approach starts by (1) ﬁnding repeating (similar) element groups that

are used across repeating units, which are constitutive of repeating

units. Then we (2) organize those elements into repeating units to

model the information structure of the infographic designs. Finally,

we (3) recognize infographic connectors and add semantic tags to

complete the structure inference that enables ﬂexible animation

arrangements.

4.1.1. Finding Similar Elements

Repeating units are designed with regularity to achieve a pleasing

visual effect. Visual elements across units are usually similar but not

exactly the same. For example, sibling shapes in an element group

may have the same size, but with different colors; sibling textboxes

may have different contents, lengths, but the same font. Identifying

them requires considerations of different similarity measurements.

Therefore, the goal of the ﬁrst step is to ﬁlter the elements that are

most probably inside the repeating units. We ﬁrst group similar ele-

ments into clusters

< C

,...,C

. For simplicity, we adopt a set

of heuristic strategies to identify similar elements for different ele-

ment types. For shapes, we extract their paths, and further categorize

the path of the shapes into a set of basic shapes [SANC17, LDH11].

Shape similarity between two visual elements is measured with the

shape type, widths, and height; Icon/image similarity is measured

with width and height; Textbox similarity is measured with font size

and textbox width (if there are multiple rows of text). Clustering

techniques (e.g., hierarchical clustering) can also be adopted.

After that, we infer the number of repeating units from the size

of clusters base on our observations from our survey: if the number

of repeating units is

, the elements across units form clusters

of similar elements of size

. Reversely, the

elements in each

cluster contribute to the N repeating units in the infographic. For

example, in Figure 5, the visual elements in the design can be put

into eight clusters based on the similarity (Figure 5a.1-3), including

six rectangles, six text boxes, six icons, etc., meaning the number

of repeating units is most probably six (

N = 6

). We compute the

size of each cluster

S = {size(C

),size(C

),...,size(C

)}

and ﬁnd

the commonest size

N = mode(S)

, meaning that many clusters have

N similar elements.

4.1.2. Identifying Repeating Units

After extracting many clusters, we assign them to repeating units

(Figure 5b.1-3). To divide them into units, we take advantage of

the regularity and proximity principle [LWL

∗

20] adopted by most

designers from our survey: (1)

Layout

: Elements are placed in

regular and harmonious patterns across units. The layout patterns

should be similar across element groups. We can merge elements

into units based on the coordinate positions if two groups have

similar layouts. For example, if similar elements are all in horizontal

Yun Wang et al. / Animated Presentation of Static Infographics with InfoMotion

Figure 5:

Structure inference process: InfoMotion ﬁrst decomposes

the visual elements in an infographic (a.1), extracts visual properties

(a.2), and clusters similar elements (a.3). Then it identiﬁes repeating

units by ﬁnding one cluster with the commonest cluster size (b.1,

N=6), merging other clusters with sizes of N (b.2) and greater than

N (b.3). Finally, it attaches layouts (c.1) and semantic tags for the

elements (c.2) and output the information structure.

layouts, we sort the elements based on horizontal positions and put

them into units one by one. (2)

Element Proximity

: Designers avoid

crossing and long distance for relative elements. Infographic designs

without standard layouts usually leverage proximity to enhance

perception – elements within a unit are more likely placed close to

each other.

The algorithm that merges clusters into units works as follows

(Algorithm 1): we ﬁrst identify the layout type of every cluster of

size(C

) = N

by classifying them into one of the four layout types

(Figure. 3). We take the most frequent layout type as the overall

unit layout (Line 3, Algorithm 1). Then we try to merge clusters of

size(C

) >= N

one by one into units by layout or proximity (Line

8-12, Algorithm 1). When the clusters’ layouts are identiﬁed and

the same with the units’ layout, and the sizes of the clusters are of

we naturally try to merge them into the

units by layout; otherwise,

we merge them by proximity. When elements are merged by layout,

they are sorted according to their positions and added to the units

accordingly. When elements are assembled by proximity (please

Algorithm 1 Merge Cluster

1: procedure MERGECLUSTER(clusters, n)

2: units ← []

3: unitLayout ← mostFrequentLayout(clusters)

4: unitCands ← ﬁlter(clusters,c ⇒ c.len ≥ n)

5: others ← ﬁlter(clusters, c ⇒ c.len < n)

6: u ← unitCands.pop()

7: while u 6= NULL do

8: layout ← detectLayout(u)

9: if layout == unitLayout and u.len == N then

10: assembleResult ← assembleByLayout(units,u)

11: else

12: assembleResult ← assembleByProximity(units,u)

13: if assembleResult == NULL then

14: add u to others

15: else

16: add assembleResult.selected to units

17: if assembleResult.le f t.len ≥ n then

18: add assembleResult.le f t to tail of unitCands

19: else

20: add assembleResult.le f t to tail of others

21: u ← unitCands.pop()

return (unitLayout, units, others)

ﬁnd more details about the algorithm of merging by proximity in

the supplementary material), we calculate the distance between

the elements in the new cluster and existing units, and put each

element in the cluster into correct units. To ensure proper merging

into units, we further leverage element regularity across units by

calculating standard deviations of the distances between current

units and the newly added elements. The standard deviation should

be lower than a threshold

to avoid grouping by mistakes. The

default value of

is set to 0.04 based on our experiments. When

the clusters are of

size(C

) > N

, we take out the elements to be

merged to the units, and check whether the remaining elements are

still of

size(C

) >= N

. It is common that the remaining elements

can still contribute to the units. We put them back to the tail of

candidate clusters. If the remaining elements are of size less than

, we put the remaining elements to the set of

others

. Finally,

the

others

elements that cannot be successfully merged into units

will be further considered as connectors, embellishments, or other

components in the next step.

4.1.3. Recognizing Connectors and Semantic Tags

After organizing elements into repeating units, we can further iden-

tify connectors between units and determine sequences. Connectors

are a special type of repeating unit. We search through visual ele-

ments of line or arrow shapes that fall in the region between any two

units. We attach the layout tags to the infographics (Figure 5c.1).

We further add semantic tags (e.g., title, icon, image, footnote, and

background, etc.) for the infographic components to enable ﬂexible

animation arrangements based on heuristic rules (Figure 5c.2). First,

we add semantic tags for the elements left out from the repeating

units and connectors as global level infographic components, like

title, description, background, and footnotes. For example, we assign

Yun Wang et al. / Animated Presentation of Static Infographics with InfoMotion

textboxes at the bottom with smaller font sizes as footnotes and

textboxes with larger font sizes as titles. Then, we also add semantic

tags to the visual elements within a unit with similar heuristics.

Usually, elements within a unit act as unit titles, unit icons, unit

background, etc. The semantic tags can be easily extended to make

the role categories ﬁner.

4.2. Presentation Arrangement

With information structure inferred, InfoMotion arranges the pre-

sentation orders and animation pacing that are applied to the visual

elements based on the design patterns summarized in our survey.

Animations should be arranged into temporal sequences logically.

Showing all animations one by one can be cluttered and unfocused,

while showing animations at the same time can be overwhelming

and hard to follow.

4.2.1. Animation Sequence

Presentation Order.

The presentation order of the animated info-

graphics are conﬁgurable parameters for users to choose. We pro-

vide three design strategies to arrange the sequence of animations.

Animation designers can choose from any of the three presenta-

tion orders to arrange the presentations: (1) Animation based on

sequences: For infographics with obvious sequence indexes, we

follow the original orders of the infographics. We go through text

boxes to check whether there are strong indicators of presentation

orders, such as number indexes (1-10) or letter indexes (A-Z). If

there are connectors and the connectors are with directions such as

arrows, we follow the directions of connectors to assign a sequence

for the units. (2) Animation sequence based on common reading

order: infographic components within an infographic design may

not have clear dependencies or logical orders. We provide choices

of reading orders (e.g., from left to right, from top to bottom, clock-

wise or counterclockwise) based on the layout of an infographic to

arrange animation sequences. (3) Animation based on semantic tags:

many designs adopt a semantic order. For example, the title appears

ﬁrst, followed by the subtitle. The footnotes appear last. For those

without a clear sequence pattern, this approach can be taken. Based

on the semantic tags, it is easy to adjust content sequences. Besides

these default choices, more sequence options can also be provided

to users.

Pacing.

Based on the two ways of arranging the elements summa-

rized in our survey, (i.e., by repeating unit, and by element group),

InfoMotion supports three pacing strategies based on our survey, i.e.,

one-by-one, all-at-once, and staggering, which can be applied to

merge animations into several stages. For example, InfoMotion can

stagger repeating units/element groups, and show elements within

units or groups all at once. By default, InfoMotion chooses “by

repeating units”, and repeating units are also shown “one by one”,

which is the most popular way from our survey. Within each unit,

visual elements are shown in a “staggering” manner with 10% over-

lapping. By setting these parameters, we can compute the sequence

and start time of showing visual elements. These parameters can

be adjusted by different application scenarios and user preferences

according to the needs.

4.2.2. Animation Effects

InfoMotion further applies animation effects for each visual element.

However, from our survey, we do not observe obvious design pat-

terns of choosing animations. Moreover, there is no consensus on

the taxonomy of animation effects. For example, animation creating

tools such as Adobe After Effects allow users to have low-level

controls over objects, while presentation tools like Keynote and

PowerPoint animate static object by injecting preset animations.

Similar to [GZL

∗

20], InfoMotion borrows the taxonomy of preset

effects used in presentation tools and tries to balance expressiveness

and conciseness.

Following this taxonomy, we identify more than 20 animation ef-

fects applied to the infographics from our dataset, including fade and

wipe that are most frequently used. Other effects like path, stretch,

swivel, and wheel appear only 1 or 2 times in our dataset. With our

dataset, we explore a data-driven method to build an element-effect

model. Through this way, we can further understand whether de-

signers have similar considerations when adopting animation effects.

We select six common animation styles, namely, fading, ﬂoating,

zooming, wiping, ﬂying, and splitting, as general choices for the

effects, covering 95.5% of the effects in the dataset. In total, we

extract 461 visual element-animation effect pairs.

We train a random forest model (with more details described in

the supplementary material) to recommend top-k ideal effects for

the visual elements [rf20]. Random forests are an ensemble learning

method of constructing a multitude of decision trees at training time

and outputting the majority vote from these individual trees as ﬁnal

predictions. The parameters for each element include element width,

element height, element shape, unit layout, and element position. We

calculate the accuracy of animation prediction if the animation effect

in one test case is included. The average accuracy of cross validation

is 86.98% (k=3), 73.69% (k=2), and 65.8% (k=1). Depending on

the ﬂexibility of animation design environments, animation design

applications can recommend top-1 or top-k animation effect(s) to the

users. When users need to modify the animation effect for a given

component, applications can recommend a ranked list of effects.

This enables users to easily modify the animation effects based on

their personal preferences.

5. Evaluation

5.1. Experiments on Structure Inference

To evaluate the infographic structure inference algorithm, we take

the animated infographics analyzed in Section 3 from real-world

designers. We manually preprocess the ﬁles to remove animations

from the source ﬁles and excluded the animations that cannot be

decomposed (e.g., videos, GIFs), adding up to 120 infographic

designs. The 120 designs contain 45.83% (55) linear, 18.33% (22)

radial, 22.50% (27) segmented, and 13.33% (16) freeform layouts.

Among them, 46.67% (56) have connectors while 53.33% (64) do

not have connectors. We have manually tagged the infographic

elements into similar element groups and repeating units. We further

tagged connectors if the infographics contain.

We compare the automatically analyzed infographic structure

(element groups and repeating units) with manually labeled ones.

Yun Wang et al. / Animated Presentation of Static Infographics with InfoMotion

(a)

(b)

Figure 6: Failure cases [sli19, pre18] from structure inference. Re-

peating units are hard to identify when (a) designers adopt uncom-

mon layout or (b) repeating unit designs are not similar enough.

The criteria for judging success are (1) the number of units are

correct, (2) the element groups containing only similar elements that

are tagged by humans. We mark the case as “correct” if elements

missed out are embellishments, meaning that the embellishments

are not repeated across units and they are very easy to be ﬁxed. From

the 120 designs, we got 94 success cases, reaching an accuracy of

78.33%. Overall, our algorithm can successfully identify infographic

structure in various design cases. We further report our success rates

regarding different aspects as follows.

• Infographic Layout.

From the layout perspective, the accuracy

of InfoMotion reaches 83.64% (46/55) for linear, 86.36% (19/22)

for radial, 62.96% (17/27) for segmented, and 75% (12/16) for

freeform layouts.

• Connector.

InfoMotion reaches an accuracy of 73.21% (41/56)

for infographics with connectors and 82.81% (53/64) for those

without connectors.

For failure cases (Figure 6), there are two common problems:

(1) The layouts do not fall into common layout types, resulting in

failures of sequencing units. For example, the textboxes in Figure

6a are placed along an arbitrary curve. The shapes and directions

cannot be parsed with InfoMotion. When infographic structures

are not correctly inferred, applications can provide interfaces that

users can inject correct structures through user interfaces to generate

animations accordingly. (2) Designers may adopt very different

visual elements across repeating units. When there is only one or

even no similar elements, it is hard to get an anchor element with the

algorithm. Consequently, it is harder to identify repeating units from

the design (e.g., Figure 6b), resulting in failure. Taking the bubbles

with different sizes and colors as an example, InfoMotion cannot

judge which ones were repeated.

5.2. Example Application

We demonstrate the utility of InfoMotion with applications in the

form of PowerPoint add-in

†

. We choose PowerPoint for three rea-

sons: (1) PowerPoint is one of the most popular presentation au-

thoring tools for general users to create infographics and present

data stories. Users can take advantage of this existing workﬂow to

generate animations and enhance their presentations automatically.

(2) A large number of static and animated infographics are designed

by designers using PowerPoint. The VSTO tools [VST20] enables

†

https://www.microsoft.com/en-us/research/project/inf

omotion/

the extraction of visual properties from the PowerPoint slides and

the attachment of animations to the infographic designs. (3) We take

the taxonomy of PowerPoint preset effects to recommend animation

effects.

InfoMotion can be implemented as a standalone application. Af-

ter computing the animations applied for different shapes, animated

infographics can be further synthesized. InfoMotion can also be inte-

grated smoothly into the workﬂow of other infographic design tools,

which can help users easily extend static designs into animated ones.

We believe InfoMotion opens up new opportunities for infographic

animation design and authoring.

Mixed-Initiative Animation Authoring.

InfoMotion can be

used to create a mixed-initiative animation authoring experience.

Users can design and import static infographics in the form of Pow-

erPoint slides. Once an infographic is designed, users can click on

the InfoMotion button to show a design panel of the animations.

Users can modify the groupings based on their needs through the

design panel (Figure 7a). They can also adjust the sequence by mov-

ing up and down the element groups. After that, users can click

the “Add animation” button on the design penal to apply animation

effects to the visual elements. They can also conﬁgure presentation

order and animation pacing through the design panel. Users can

click “Play” or “Preview” button in PowerPoint to review the ani-

mation effects. If they want to modify the animations, they can also

leverage the InfoMotion Designer to select a group of objects on

canvas. As such, users can easily select elements and ﬁne-tune the

animation conﬁgurations at any granularity. Integrating InfoMotion

in a mixed-initiative authoring tool preserves human creativity but

eases the repetitive efforts in creating animations.

Animation Recommendation.

Another possible usage of Info-

Motion is to recommend many animated infographic designs based

on one static infographic. Users no longer need to imagine anima-

tion effects before they start to design. They just need to choose

the animations that they prefer most. This can reduce the barrier of

using animations, which can be especially useful for novice users.

As shown in Figure 7b, when a user completes the design of an

infographic, with just one click, InfoMotion shows a list of anima-

tion design candidates. The design candidates are generated through

the combinations of different animation settings, such as animation

sequence, staging, and effects. Users can choose one from the can-

didates and further adjust the animation designs to ﬁne tune the

results.

5.3. User Study

We conduct a user study to gather subjective feedback and under-

stand the quality of the InfoMotion-generated animated infograph-

ics, by comparing them with the designer-crafted ones randomly

selected from the animated infographic tutorials in our survey.

Participants.

We recruited 32 participants (19 male, 13 female,

aged from 23 to 52) from a local tech company. The participants

include data analysts, researchers, engineers, program managers,

and intern students. They are general users who need to present data

and communicate insights through slide decks in their daily work.

All of them have used presentation tools like PowerPoint or Keynote

to create animated presentations. Five of them had experience with

Yun Wang et al. / Animated Presentation of Static Infographics with InfoMotion

(a) (b)

Figure 7:

Sample applications: (a) A PowerPoint add-in that supports mixed-initiative animation authoring; (b) Another add-in that supports

animation recommendation for one static infographic. It provides multiple animation design candidates for users to choose.

graphic design tools such as Adobe Illustrator and Photoshop. No

participant reported vision impairments.

Study Design.

To avoid users paying attention to the graphical

designs of the infographics, we conduct a pair-wise within-subject

comparison study between two groups of animations. (1) Auto-

generated group: With the prototype applications (Section 5.2) at

hand, we prepare the study stimuli by randomly selecting info-

graphic templates in the form of PowerPoint ﬁles, which might

indicate the high-quality of the animated infographics. Then we

manually remove all the animation actions, resulting in 18 static

infographic designs. We run InfoMotion for each static infographic

and pick the top-ranked generated animated infographic for the

study. (2) Designer-crafted group: the original animated infograph-

ics corresponding to the auto-generated animations naturally form

the designer-crafted group. We calculate the number of mouse clicks

in the tutorial videos that demonstrate the authoring process of the

animations without design iterations. The average number is 73.6

clicks (max = 176, min = 13, median = 56), excluding the clicks

for static infographic design, which could be more difﬁcult if the

resulting designs are indeﬁnite, and the users are non-experts. This

indicates the complexity of the animation design tasks and the ef-

forts involved to author these animations from static infographic

design.

Procedure.

The participants rated their preferences between the

two groups of animations with a 7-point bidirectional Likert-scale.

Each participant performed 18 trials. In each trial, the participants

were shown a pair of animated infographics. The sequence of the

two animations and the 18 pairs was shufﬂed and users were not

informed that half of them are auto-generated until they completed

rating. The participants were asked to view pairs of animations by

clicking the “preview” button for animations when viewing a Power-

Point ﬁle. For each pair, we used 7-point bidirectional Likert-Scale

ranging from “the ﬁrst one is strongly better” to “the second one

is strongly better” to understand users’ preference between the two

animations. During the study, participants were encouraged to think

aloud or write down their reasons for choices. After the completion

of the ratings, the participants were told that half of the animations

are automatically generated. We conducted a semi-structured inter-

view to review and discuss the considerations involved during their

ratings and asked them to complete a short demographic survey.

Subjective Ratings.

All participants completed the ratings with

roughly 10-15 minutes in total. We collected a total of

32×18 = 576

ratings. The median of all ratings is

with the scale

[−3,3]

. Figure

8 shows the percentages and absolute numbers of subjective ratings

ranging from -3 to 3 provided by the participants for each animation

pair (AP). Both the median values and the p-values obtained from

the Wilconxon test on whether the medians equal to 0 are also listed

for each pair. After applying Bonferroni correction to set the signiﬁ-

cant value at

c = 0.05/18

, 13 out of the 18 tests yield non-signiﬁcant

p-values (

p > c

), which indicates that participants shown no prefer-

ence over the designer-crafted and machine-generated animations

in the majority of the trials. The participants showed statistically

signiﬁcant preference on ﬁve human-generated infographics (AP1,

AP4, AP9, AP10, AP15). A detailed investigation over those pairs

suggests that humans perform better than machines when complex

or context-dependent animations are required. For example, the

main body of one infographic (AP10) is a cartoon pencil made up

of multiple rectangles. While the auto-generated animation shows

the pencil segment by segment, the designer-crafted one shows the

pencil as a whole. The other examples (AP1, AP4, AP9, AP15) uses

multiple shapes to mimic layers, shadows, and depth of real-world

objects.

Qualitative Observations.

The overall reactions from the partic-

ipants were promising. Most participants felt it is more engaging

to watch animations than static infographics. As P9 said, “I think

the difference between these pairs is not much. All of them are much

better than static ones.” Participants did not realize that half of the

cases were machine-generated. After knowing that half of them were

automatically generated, many of the participants felt surprised, with

one saying “Can you let me know which ones are automatically

generated? It is not easy to guess.” Another participant mentioned,

“Hard to imagine they are generated with just one click!”

Through the interview, we also learn lessons from users’ prefer-

ences of animation designs. First, users have different preferences

of animation ordering. Some of them preferred to adopt a common

reading order (e.g., left-to-right, up-to-down), while some others

Yun Wang et al. / Animated Presentation of Static Infographics with InfoMotion

-100% 60%

Percentage

Strongly

Disagree

Disagree Somewhat

Disagree

Neutral Somewhat

Agree

Agree Strongly

Agree

Response

80%40%20%0%-20%-40%-60%-80%

[M=-1.0, P<0.001]

[M= 0.0, P=0.640]

[M= 1.5, P=0.034]

[M=-2.0, P=0.001]

[M=-0.5, P=0.128]

[M=-0.5, P=0.843]

[M= 0.0, P=0.036]

[M= 1.0, P=0.025]

[M=-1.0, P<0.001]

[M=-2.0, P<0.001]

[M= 0.5, P=0.764]

[M= 1.5, P=0.006]

[M= 1.0, P=0.008]

[M= 1.0, P=0.012]

[M=-2.0, P=0.003]

[M= 0.0, P=0.793]

[M= 0.0, P=0.786]

[M= 0.0, P=0.166]

AP 1

AP 2

AP 3

AP 4

AP 5

AP 6

AP 7

AP 8

AP 9

AP 10

AP 11

AP 12

AP 13

AP 14

AP 15

AP 16

AP 17

AP 18

Figure 8:

User ratings for the 18 animations pairs for each in-

fographic design: from strongly prefer designer-crafted ones to

strongly prefer auto-generated ones.

prefer to show the theme ﬁrst. For example, P2 preferred up-to-

down order because “it is easy to follow with my eyes”, while P18

preferred theme-ﬁrst order because “I prefer (unit) titles shown ﬁrst.”

This indicates the importance of applications to have a user interface

for users inject their preferences. Another interesting ﬁnding is the

usage of redundancy. Some designers may highlight elements by

ﬂickering or bouncing for a short time, applying more than one

animation effects to the same shape to emphasize some parts of the

infographics, while the automatic-generated animations apply one

animation for each visual element by default. The redundancy makes

some of the human-designed animations lively and attractive, while

machine-generated ones seem a bit rigid. Some participants (2/32)

thought the animations (designed by humans) are nonsensical and

too superﬂuous. One participant said, “This one should be reduced.

I don’t want to watch this bubble ﬂickering for such a long time”,

while some other participants (2/32) mentioned these animations

were interesting and attractive.

6. Discussion

Beneﬁts and Drawbacks of Automatic Generation

The creation

of animated infographics is time-consuming even for professional

designers. Users need to take a great amount of time assigning

timings and effects one by one to each visual element. InfoMo-

tion can save these manual efforts and make it possible to generate

delicate animations in one click even for non-designers. However,

the automatically generated animations are not perfect and cannot

be as creative as the animations made by top designers. For exam-

ple, designers sometimes design more creative animations based on

physical properties (e.g., balls falling under gravity, clouds ﬂoating

in the air, papers tearing apart). Current design of InfoMotion only

considers abstract shapes. To make these designs possible, we need

to further understand graphical objects with real-world meanings to

make animations more attractive and natural.

Animation Designing vs. Authoring

Most common visualiza-

tion authoring tools assume that users want to author, instead of

design a visualization [SLR

∗

19]. In other words, the authoring tools

assume users have a speciﬁc design in mind or sketched on paper.

However, it is very hard for users to sketch animations on paper.

With traditional animation authoring tools, users have to imagine

in mind, which restricts the exploration of animation design space.

As illustrated in Section 5.2, the automatic method has changed the

authoring process of animation design in two ways. With just a few

clicks, users can start from many draft animations. They can even

easily compare among multiple design prototypes in parallel and

get inspiration from the generated designs.

Animated Infographic Design Paradigms

The current design

of the animation generation considers the direct generation of in-

fographics from static ones. To better exploit human ingenuity and

creativity, other design paradigms can also be explored. Supported

by our information model, we can consider a 1-to-many way – to

transfer the animation design from one unit to another, or even one

infographic to another infographic with similar layouts and struc-

tures. This approach leaves out the labor-intensive parts but injects

human’s creativity to the creation of animations.

Limitations

Based on the repeating units, InfoMotion does not

support more complex and advanced infographic-style diagrams,

visualizations, and tables that encode size, length, position, and color

of the units with data ﬁelds. The encoding scheme varies across chart

types. For example, pyramid charts have different colors and shapes

for each unit. Gantt charts may use color and length to encode

numerical value or duration. Therefore, the repeating patterns may

not be easy to detect without further knowledge of chart types.

Another limitation is the number of cases we adopt for evaluation.

Although a number of animated infographic videos and GIFs could

be collected, source ﬁles that can be converted to vector images are

not easy to gain. In the future, we plan to invite human users to

produce more animations from the static designs for more detailed

comparison analysis. Similarly, the performance of the decision tree

model is highly related to the animations we collect. We hope to

collect more animations to improve the model in the future.

7. Conclusion and Future Work

In this paper, we contribute an automated approach to present an-

imated infographics from static infographic designs. Through an

analysis of existing animated infographic designs, we summarize

common patterns in real-world design practices. Based on the anal-

ysis, InfoMotion can model the information structures, compose

the visual elements into animation sequences, and apply a set of

animation effects. To illustrate the usage of InfoMotion, we develop

example applications in the form of PowerPoint add-ins, which au-

tomatically attaches various built-in animation effects to the visual

elements in a slide. We further conduct a user study to compare ani-

mations generated by InfoMotion with the ones crafted by designers.

Our study shows the results of automatic generated infographic

designs are promising in terms of animation quality.

We believe InfoMotion opens new avenues and encourages more

research on the authoring and generation of animated infographics.

In the future, we plan to develop an animated infographic platform

that can take multiple infographic design forms as input and help

users generate and publish enjoyable infographic videos. We also

plan to generate and recommend animated infographics directly

from datasets or text information based on advanced infographic

generation techniques.

Yun Wang et al. / Animated Presentation of Static Infographics with InfoMotion

References

[6op19] 6 options infographic slide in powerpoint / business presentation

/ powerpoint tutorial.

https://www.youtube.com/watch?v=LGC9D

A2xDR4, 2019. 4

[9op19]

9 options infographic slide in powerpoint / animated powerpoint

slide design tutorial.

https://www.youtube.com/watch?v=BSf-b

3puhCA, 2019. 4

[ARL

∗

15]

AMINI F., RICHE N. H., LEE B., HURTER C., IRANI

P.: Understanding data videos: Looking at narrative visualization

through the cinematography lens. In Proceedings of the 33rd An-

nual ACM Conference on Human Factors in Computing Systems, CHI

2015, Seoul, Republic of Korea, April 18-23, 2015 (2015), pp. 1459–

1468. URL:

https://doi.org/10.1145/2702123.2702431

doi:

10.1145/2702123.2702431. 2

[ARL

∗

17]

AMINI F., RICHE N. H., LEE B., MONROY-HERNÁNDEZ A.,

IRANI P.: Authoring data-driven videos with dataclips. IEEE Trans. Vis.

Comput. Graph. 23, 1 (2017), 501–510. URL:

https://doi.org/10.

1109/TVCG.2016.2598647

doi:10.1109/TVCG.2016.2598647

. 1,

[bal18]

How to design easy and beautiful slide in powerpoint - part 2 |

balance scale design.

https://www.youtube.com/watch?v=X66-F

HjZE4U, 2018. 4

[BKO

∗

17]

BYLINSKII Z., KIM N. W., O’DONOVAN P., ALSHEIKH S.,

MADAN S., PFISTER H., DURAND F., RUSSELL B., HERTZMANN A.:

Learning visual importance for graphic designs and data visualizations.

In Proceedings of the 30th Annual ACM symposium on user interface

software and technology (2017), pp. 57–69. 2

[BMG

∗

10]

BATEMAN S., MANDRYK R. L., GUTWIN C., GENEST A.,

MCDINE D., BROOKS C.: Useful junk? the effects of visual embellish-

ment on comprehension and memorability of charts. In Proceedings of

the SIGCHI conference on human factors in computing systems (2010),

pp. 2573–2582. 2

[BPF14]

BACH B., PIETRIGA E., FEKETE J.: Visualizing dynamic net-

works with matrix cubes. In CHI Conference on Human Factors in

Computing Systems, CHI’14, Toronto, ON, Canada - April 26 - May

01, 2014 (2014), pp. 877–886. URL:

https://doi.org/10.1145/

2556288.2557010, doi:10.1145/2556288.2557010. 2

[BS90]

BAECKER R., SMALL I.: Animation at the interface. The art of

human-computer interface design (1990), 251–267. 2

[BVB

∗

13]

BORKIN M. A., VO A. A., BYLINSKII Z., ISOLA P.,

SUNKAVALLI S., OLIVA A., PFISTER H.: What makes a visualization

memorable? IEEE Transactions on Visualization and Computer Graphics

19, 12 (Dec 2013), 2306–2315. doi:10.1109/TVCG.2013.234. 2

[CDBF10]

CHEVALIER F., DRAGICEVIC P., BEZERIANOS A., FEKETE

J.: Using text animated transitions to support navigation in document

histories. In Proceedings of the 28th International Conference on Human

Factors in Computing Systems, CHI 2010, Atlanta, Georgia, USA, April

10-15, 2010 (2010), pp. 683–692. URL:

https://doi.org/10.1145/

1753326.1753427, doi:10.1145/1753326.1753427. 2

[con19]

4 step infographic concept slide for powerpoint - free ppt.

https:

//www.youtube.com/watch?v=hmApxMVNg10, 2019. 4

[CRP

∗

16]

CHEVALIER F., RICHE N. H., PLAISANT C., CHALBI A.,

HURTER C.: Animations 25 years later: New roles and opportunities.

In Proceedings of the International Working Conference on Advanced

Visual Interfaces, AVI 2016, Bari, Italy, June 7-10, 2016 (2016), pp. 280–

287. URL:

https://doi.org/10.1145/2909132.2909255

doi:

10.1145/2909132.2909255. 1, 2, 3

[CRP

∗

20]

CHALBI A., RITCHIE J., PARK D., CHOI J., ROUSSEL N.,

ELMQVIST N., CHEVALIER F.: Common fate for animated transitions in

visualization. IEEE Trans. Vis. Comput. Graph. 26, 1 (2020), 386–396.

URL:

https://doi.org/10.1109/TVCG.2019.2934288

doi:10.

1109/TVCG.2019.2934288. 2

[CWW

∗

19]

CHEN Z., WANG Y., WANG Q., WANG Y., QU H.: Towards

automated infographic design: Deep learning-based auto-extraction of

extensible timeline. IEEE transactions on visualization and computer

graphics 26, 1 (2019), 917–926. 2

[CZW

∗

20]

CUI W., ZHANG X., WANG Y., HUANG H., CHEN B., FANG

L., ZHANG H., LOU J., ZHANG D.: Text-to-viz: Automatic gener-

ation of infographics from proportion-related natural language state-

ments. IEEE Trans. Vis. Comput. Graph. 26, 1 (2020), 906–916. URL:

https://doi.org/10.1109/TVCG.2019.2934785

doi:10.1109/

TVCG.2019.2934785. 2

[DBJ

∗

11]

DRAGICEVIC P., BEZERIANOS A., JAVED W., ELMQVIST N.,

FEKETE J.: Temporal distortion for animated transitions. In Proceedings

of the International Conference on Human Factors in Computing Systems,

CHI 2011, Vancouver, BC, Canada, May 7-12, 2011 (2011), pp. 2009–

2018. URL:

https://doi.org/10.1145/1978942.1979233

doi:

10.1145/1978942.1979233. 2

[DCZL15]

DU F., CAO N., ZHAO J., LIN Y.: Trajectory bundling for ani-

mated transitions. In Proceedings of the 33rd Annual ACM Conference on

Human Factors in Computing Systems, CHI 2015, Seoul, Republic of Ko-

rea, April 18-23, 2015 (2015), pp. 289–298. URL:

https://doi.org/

10.1145/2702123.2702476, doi:10.1145/2702123.2702476. 2

[DMM04]

DESOLNEUX A., MOISAN L., MOREL J.-M.: Gestalt theory

and computer vision. In Seeing, Thinking and Knowing. Springer, 2004,

pp. 71–101. 3

[EDF08]

ELMQVIST N., DRAGICEVIC P., FEKETE J.: Rolling the dice:

Multidimensional visual exploration using scatterplot matrix navigation.

IEEE Trans. Vis. Comput. Graph. 14, 6 (2008), 1539–1148. URL:

https://doi.org/10.1109/TVCG.2008.153

doi:10.1109/TVCG

.2008.153. 2

[Fis10]

FISHER D.: Animation for visualization: Opportunities and draw-

backs., 2010. 1

[ﬂo20] Flourish. https://flourish.studio, 2020. 3

[FWD

∗

19]

FU X., WANG Y., DONG H., CUI W., ZHANG H.: Visualiza-

tion assessment: A machine learning approach. In 2019 IEEE Visualiza-

tion Conference (VIS) (2019), IEEE, pp. 126–130. 2

[GZL

∗

20]

GE T., ZHAO Y., LEE B., REN D., CHEN B., WANG Y.: Canis:

A high-level language for data-driven chart animations. In Computer

Graphics Forum (2020), vol. 39, Wiley Online Library, pp. 607–617. 2, 7

[HKF15]

HAROZ S., KOSARA R., FRANCONERI S. L.: Isotype visualiza-

tion: Working memory, performance, and engagement with pictographs.

In Proceedings of the 33rd annual ACM conference on human factors in

computing systems (2015), pp. 1191–1200. 2

[HKL17]

HULLMAN J., KOSARA R., LAM H.: Finding a clear path:

Structuring strategies for visualization sequences. In Computer Graphics

Forum (2017), vol. 36, Wiley Online Library, pp. 365–375. 4

[HR07]

HEER J., ROBERTSON G. G.: Animated transitions in statistical

data graphics. IEEE Trans. Vis. Comput. Graph. 13, 6 (2007), 1240–

1247. URL:

https://doi.org/10.1109/TVCG.2007.70539

doi:

10.1109/TVCG.2007.70539. 2

[HVF13]

HURON S., VUILLEMOT R., FEKETE J.: Visual sedimentation.

IEEE Trans. Vis. Comput. Graph. 19, 12 (2013), 2446–2455. URL:

https://doi.org/10.1109/TVCG.2013.227

doi:10.1109/TVCG

.2013.227. 2

[JKS

∗

17]

JUNG D., KIM W., SONG H., HWANG J.-I., LEE B., KIM B.,

SEO J.: Chartsense: Interactive data extraction from chart images. In

Proceedings of the 2017 chi conference on human factors in computing

systems (2017), pp. 6706–6717. 3

[KCH19]

KIM Y., CORRELL M., HEER J.: Designing animated tran-

sitions to convey aggregate operations. Comput. Graph. Forum 38, 3

(2019), 541–551. URL:

https://doi.org/10.1111/cgf.13709

doi:10.1111/cgf.13709. 2

[KH20]

KIM Y., HEER J.: Gemini: A grammar and recommender

system for animatedtransitions in statistical graphics. arXiv preprint

arXiv:2009.01429 (2020). 2

Yun Wang et al. / Animated Presentation of Static Infographics with InfoMotion

[KSL

∗

17]

KIM N. W., SCHWEICKART E., LIU Z., DONTCHEVA M., LI

W., POPOVIC J., PFISTER H.: Data-driven guides: Supporting expressive

design for information graphics. IEEE Transactions on Visualization and

Computer Graphics 23, 1 (Jan 2017), 491–500.

doi:10.1109/TVCG

.2016.2598620. 2

[LDH11]

LI D., DENG L., HUANG Z.: Shape similarity computation for

svg. International Journal of Computational Science and Engineering 6,

1-2 (2011), 105–113. 5

[LTW

∗

18]

LIU Z., THOMPSON J., WILSON A., DONTCHEVA M., DE-

LOREY J., GRIGG S., KERR B., STASKO J.: Data illustrator: Augmenting

vector design tools with lazy data binding for expressive visualization

authoring. In Proceedings of the 2018 CHI Conference on Human Factors

in Computing Systems (New York, NY, USA, 2018), CHI ’18, ACM,

pp. 123:1–123:13. URL:

http://doi.acm.org/10.1145/3173574.

3173697, doi:10.1145/3173574.3173697. 2

[LWL

∗

20]

LU M., WANG C., LANIR J., NANXUAN Z., HANSPETER

P., COHEN-OR D., HUANG H.: Exploring visual information ﬂows in

infographics. In Proceedings of the 2020 CHI Conference on Human

Factors in Computing Systems (New York, NY, USA, 2020), CHI ’20,

ACM. URL:

http://doi.acm.org/10.1145/3173574.3173697

doi:https://doi.org/10.1145/3313831.3376263. 2, 3, 5

[map18]

Powerpoint animation tutorial - map location.

https://www.

youtube.com/watch?v=-tP7w7Jju5Y, 2018. 4

[mil18]

Timeline, process, milestones, achievements, targets, sales, steps,

workﬂow design in powerpoint.

https://www.youtube.com/watc

h?v=ej_uDR6qgSQ, 2018. 4

[pen19]

How to make a creative 3d pencil infographic on powerpoint

| powerpoint show.

https://www.youtube.com/watch?v=aqaSt_

61wp4, 2019. 4

[PH17] POCO J., HEER J.: Reverse-engineering visualizations: Recover-

ing visual encodings from chart images. In Computer Graphics Forum

(2017), vol. 36, Wiley Online Library, pp. 353–363. 3

[pre18]

Powerpoint presentation slide animation tutorial.

https://www.

youtube.com/watch?v=QF9kG8QTgVI, 2018. 8

[pro19]

Process slide in powerpoint.

https://www.youtube.com/wa

tch?v=ND3WY2FSrl8, 2019. 4

[rf20]

sklearn.ensemble.randomforestclassiﬁer.

https://scikit-lea

rn.org/stable/modules/generated/sklearn.ensemble.Rando

mForestClassifier.html, 2020. 7

[RLB19]

REN D., LEE B., BREHMER M.: Charticulator: Interactive

construction of bespoke chart layouts. IEEE Transactions on Visualization

and Computer Graphics 25, 1 (Jan 2019), 789–799.

doi:10.1109/TV

CG.2018.2865158. 2

[RM15]

RUCHIKACHORN P., MUELLER K.: Learning visualizations

by analogy: Promoting visual literacy through visualization morphing.

IEEE transactions on visualization and computer graphics 21, 9 (2015),

1028–1044. 2

[rou19]

4 steps infographic slide in powerpoint | round squares infograph-

ics. https://www.youtube.com/watch?v=Zq9TdHJa2UI, 2019. 4

[SANC17]

SEEL-AUDOM C., NAIYAPO W., CHOUVATUT V.: A search

for geometric-shape objects in a vector image: Scalable vector graphics

(svg) ﬁle format. In 2017 9th International Conference on Knowledge

and Smart Technology (KST) (2017), IEEE, pp. 305–310. 5

[SHL

∗

16]

SIEGEL N., HORVITZ Z., LEVIN R., DIVVALA S., FARHADI

A.: Figureseer: Parsing result-ﬁgures in research papers. In European

Conference on Computer Vision (2016), Springer, pp. 664–680. 3

[SKC

∗

11]

SAVVA M., KONG N., CHHAJTA A., FEI-FEI L., AGRAWALA

M., HEER J.: Revision: Automated classiﬁcation, analysis and redesign

of chart images. In Proceedings of the 24th annual ACM symposium on

User interface software and technology (2011), pp. 393–402. 3

[sli19]

Timeline slide in powerpoint.

https://www.youtube.com/wa

tch?v=hdhxNqE5pKg, 2019. 8

[SLR

∗

19]

SATYANARAYAN A., LEE B., REN D., HEER J., STASKO J.,

THOMPSON J., BREHMER M., LIU Z.: Critical reﬂections on visualiza-

tion authoring systems. IEEE transactions on visualization and computer

graphics 26, 1 (2019), 461–471. 10

[tim18]

Timeline infographic - animated powerpoint template.

https:

//www.youtube.com/watch?v=R-gXjpNtZHs, 2018. 4

[tim19]

Timeline slide in powerpoint.

https://www.youtube.com/wa

tch?v=hdhxNqE5pKg, 2019. 4

[TLLS20]

THOMPSON J., LIU Z., LI W., STASKO J.: Understanding

the design space and authoring paradigms for animated data graphics.

In Computer Graphics Forum (2020), vol. 39, Wiley Online Library,

pp. 207–218. 2, 5

[TMB02]

TVERSKY B., MORRISON J. B., BETRANCOURT M.: Anima-

tion: can it facilitate? International journal of human-computer studies

57, 4 (2002), 247–262. 2

[tut17]

Animated powerpoint infographic slide design tutorial.

https:

//www.youtube.com/watch?v=OZKTzpbnGOA, 2017. 4

[Vis20] Visme. https://www.visme.co/, 2020. 3

[VST20]

Vsto add-in developer’s guide.

https://docs.microsoft

.com/en-us/office/dev/add-ins/overview/learning-path-t

ransition, 2020. 8

[WCL

∗

16]

WANG Y., CHEN Z., LI Q., MA X., LUO Q., QU H.: Ani-

mated narrative visualization for video clickstream data. In SIGGRAPH

Asia 2016 Symposium on Visualization (2016), pp. 1–8. 2

[WSK

∗

19]

WANG Y., SEGAL A., KLATZKY R., KEEFE D. F., ISENBERG

P., HURTIENNE J., HORNECKER E., DWYER T., BARRASS S.: An

emotional response to the value of visualization. IEEE computer graphics

and applications 39, 5 (2019), 8–17.

[WSZ

∗

20]

WANG Y., SUN Z., ZHANG H., CUI W., XU K., MA X.,

ZHANG D.: Datashot: Automatic generation of fact sheets from tabular

data. IEEE Trans. Vis. Comput. Graph. 26, 1 (2020), 895–905. URL:

https://doi.org/10.1109/TVCG.2019.2934398

doi:10.1109/

TVCG.2019.2934398. 2

[WTD

∗

20]

WU A., TONG W., DWYER T., LEE B., ISENBERG P., QU H.:

Mobilevisﬁxer: Tailoring web visualizations for mobile phones leveraging

an explainable reinforcement learning framework. IEEE Transactions on

Visualization and Computer Graphics (2020). 3

[WWS

∗

21]

WU A., WANG Y., SHU X., MORITZ D., CUI W., ZHANG

H., ZHANG D., QU H.: Survey on artiﬁcial intelligence approaches for

visualization data. arXiv preprint arXiv:2102.01330 (2021). 2

[WXL

∗

21]

WU A., XIE L., LEE B., WANG Y., CUI W., QU H.: Learn-

ing to automate chart layout conﬁgurations using crowdsourced paired

comparison. arXiv preprint arXiv:2101.03680 (2021). 2

[WZH

∗

18]

WANG Y., ZHANG H., HUANG H., CHEN X., YIN Q., HOU

Z., ZHANG D., LUO Q., QU H.: Infonice: Easy creation of information

graphics. In Proceedings of the 2018 CHI Conference on Human Factors

in Computing Systems (New York, NY, USA, 2018), CHI ’18, ACM,

pp. 335:1–335:12. URL:

http://doi.acm.org/10.1145/3173574.

3173909, doi:10.1145/3173574.3173909. 2

[XHRC

∗

18]

XIA H., HENRY RICHE N., CHEVALIER F., DE ARAUJO

B., WIGDOR D.: Dataink: Direct and creative data-oriented drawing. In

Proceedings of the 2018 CHI Conference on Human Factors in Computing

Systems (2018), ACM, p. 223. 2

[ZCL18]

ZHAO N., CAO Y., LAU R. W.: What characterizes personalities

of graphic designs? ACM Transactions on Graphics (TOG) 37, 4 (2018),

1–15. 2