Posted: February 26th, 2022

Mark Embellishments Paper Report

Rules of Visual Encoding

No unjus3fied 3D: Power of the plane

•  high-ranked spa3al posi3on channels: planar spa3al posi3on – not depth!

No unjus3fied 3D: Danger of depth •  we don’t really live in 3D: we see in 2.05D •  acquire more info on image plane quickly from eye movements –acquire more info for depth slower, from head/body mo3on

Occlusion hides informa3on

•  Occlusion •  interac3on complexity

[Distor(onViewingTechniques for 3D Data.Carpendale et al.InfoVis1996.]

Perspec3ve distor3on loses informa3on

•  perspec3ve distor3on – interferes with all size channel encodings – power of the plane is lost!

[Visualizing the Results of Mul(mediaWeb Search Engines. Mukherjea, Hirata, and Hara. InfoVis 96]

Tilted text isn’t legible

•  text legibility •  far worse when 3lted from image plane

[Visualizing the World- Wide Web with the Naviga(onal View Builder. Mukherjea and Foley. Computer Networks and ISDN Systems, 1995.]

No unjus3fied 3D example : Timeseries data

•  extruded curves: detailed comparisons impossible

[Cluster and Calendar based Visualiza(on of Time Series Data. van Wijk and van Selow, Proc. InfoVis 99.]

No unjus3fied 3D example : Transform for new data abstrac3on

•  derived data: cluster hierarchy •  juxtapose mul3ple views: calendar, superimposed 2D curves

Jus3fied 3D: shape percep3on

•  benefits outweigh costs when task is shape percep3on for 3D spa3al data

•  interac3ve naviga3on supports synthesis across many viewpoints

[Image-Based Streamline Genera(on and Rendering. Li and Shen. IEEE Trans. Visualiza(on and Computer Graphics (TVCG) 13:3 (2007), 630–640.]

No unjus3fied 3D •  3D legi3mate for true 3D spa3al data •  3D needs very careful jus3fica3on for abstract data –  enthusiasm in 1990s, but now skep3cism –  be especially careful with 3D for point clouds or networks

[WEBPATH-a three dimensional Web history. Frecon and Smith. Proc. InfoVis 1999]

No unjus3fied 2D •  consider whether network data requires 2D spa3al layout –  especially if reading text is central to task!

–  arranging as network means lower informa3on density and harder label lookup compared to text lists

•  benefits outweigh costs when topological structure/context important for task –  be especially careful for search results, document collec3ons, ontologies

Eyes beat memory •  principle: external cogni3on vs. internal memory

–  easy to compare by moving eyes between side-by-side views –  harder to compare visible item to memory of what you saw

•  implica3ons for anima3on –  great for choreographed storytelling –  great for transi3ons between two states –  poor for many states with changes everywhere

•  consider small mul3ples instead

Eyes beat memory example: Cerebral

•  small mul3ples: one graph instance per experimental condi3on –  same spa3al layout –  color differently, by condi3on

[Cerebral:Visualizing Mul(ple Experimental Condi(ons on a Graph with Biological Context. Barsky, Munzner, Gardy, and Kincaid. IEEE Trans. Visualiza(on and Computer Graphics (Proc. InfoVis 2008 – Affordable Custom Essay Writing Service | Write My Essay from Pro Writers) 14:6 (2008 – Affordable Custom Essay Writing Service | Write My Essay from Pro Writers), 1253–1260.]

Why not anima3on? •  disparate frames and

regions: comparison difficult –  vs con3guous frames –  vs small region –  vs coherent mo3on of group

•  change blindness –  even major changes difficult to no3ce if mental buffer wiped

•  safe special case –  animated transi3ons

Resolu3on beats immersion •  immersion typically not helpful for abstract data

–  do not need sense of presence or stereoscopic 3D •  resolu3on much more important

–  pixels are the scarcest resource –  desktop also be[er for workflow integra3on

•  virtual reality for abstract data very difficult to jus3fy

[Development of an informa(on visualiza(on tool using virtual reality. Kirner and Mar(ns. Proc. Symp.Applied Compu(ng 2000]

Overview first, zoom and filter, details on demand

•  influen3al mantra from Shneiderman [The Eyes HaveIt:ATask by DataTypeTaxonomy for Informa(onVisualiza(ons. Shneiderman. Proc. IEEE Visual Languages, pp. 336–343, 1996.]

•  overview = summary – microcosm of full vis design problem

•  Nuances – beyond just two levels: mul3-scale structure – difficult when scale huge: give up on overview and browse local neighborhoods?

[Search, Show Context, Expand on Demand: Suppor(ng Large Graph Explora(on with Degree-of-Interest. van Ham and Perer.IEEETrans.Visualiza(on and Computer Graphics (Proc.InfoVis 2009) 15:6 (2009), 953–960.]

Func3on first, form next

•  start with focus on func3onality – straigh`orward to improve aesthe3cs later on, as refinement

– if no exper3se in-house, find good graphic designer to work with

•  dangerous to start with aesthe3cs – usually impossible to add func3on retroac3vely

Artery Visualiza3ons for Heart Disease Diagnosis

HemoViz: Design study + evalua3on

•  forma3ve study with experts –  task taxonomy

•  HemoViz design •  deploy a[empt fails

–  experts balk: demand 3D and rainbows

•  quan3ta3ve user study –  med students, real data –  91% with 2D/diverging vs 39% with 3D/ rainbows

–  experts willing to use

[Fig 1. Borkin et al. Artery Visualiza(ons for Heart Disease Diagnosis. Proc InfoVis 2011.]]

Study Results: Error

Study Results: Time

Cri3que •  many strengths

–  careful and well jus3fied design, convincing human-subjects experiment • bringing visualiza3on best prac3ces to medical domain

•  Limita3on –  paper does not clearly communicate why colormap is diverging not sequen3al

•  answer by email •  doctors care about extremely high and extremely low ESS (scalar) values

–  high values (top of scale, dark grey): extreme blood flow pa[erns may relate to heart malfunc3ons – but not imminently life threatening and don’t indicate plaque loca3ons

–  low values (bo[om of scale, dark red): very diseased regions with lots of plaque, docs care a lot! – much debate from doctors on where is boundary between “normal” and “low” ESS values •  most think below 3 Pa are indica3ve of disease but many argue other values in the 2-4 range. •  all docs agree that values below 2 Pa are increasingly dangerous disease levels. •  thus map has transi3on at 3 Pa for the diverging point and truly red below 2 Pa

•  why con3nuous not segmented? –  doctors gain tremendous insight by seeing the subtle pa[erning of the ESS values –  par3cularly varying values in red region – pa[erns help them understand disease progression

and severity •  especially useful for deciding what types of interven3ons to prescribe for the pa3ent

Further Reading

•  Exploring and Reducing the Effects of Orienta(on on Text Readability in Volumetric Displays. Grossman et al. CHI 2007

Order | Check Discount