Instructor: Guoning
Chen
Office: PGH
566
Email: chengu at cs . uh . edu
Lecture
time: Tue/Thur
1~2:30PM
Location: M 113
Office Hours: Tu/Th 2:30-3:30PM or by appointment (PLEASE send me an email to schedule additional appointments BEFORE coming to my office!!!)
TA: Lieyu Shi (shilieyu91@gmail.com)
Office hours: PGH 313, Mon/Wed 1-3pm
Our visual perception channel can perceive and process information much faster than other channels. We can easily identify the difference and/or patterns in the data if a proper visual representation of the data is presented. This is what visualization is aiming to achieve, i.e., converting or mapping data to its proper visual representation to enable knowledge discovery and aid decision making. Visualization is an active and popular direction in data science research, AND an enabling technique for almost every real-world application. Techniques from computer science, mathematics, cognitive and perception science, and human-computer interaction are often adapted for various visualization problems.
This introductory course covers topics from a number of sub-fields in visualization and aims to show students how data visualization can be enormously important in many real-world applications. Through this course, the students are expected to (1) get familiar with a number of important concepts and techniques/methods for the visualization of different basic types of data, (2) foster the ability to select the proper visualization techniques when given a practical data visualization problem. This course serves as one of the core introductory level graduate courses. Together with the existing courses on imaging and computer graphics, it helps build a complete course catalog in the direction of visual computing.
You will be required to accomplish a number of programming assignments, implementing a number of classic visualization techniques introduced in the class. To ease your implementation, a C++ skeleton code is provided. If you want to use a different language to finish your assignments, you will need to convert the provided C++ skeleton code to the language of your choice.
Visualization techniques are highly application dependent and highly diversified! There is currently no a good book that can summarize all available techniques. However, the following textbooks provide a good introduction to some well-established techniques for a number of fundamental visualization problems.
Project assignments – 45%;
Mid-term exam and final open-note quiz – 25%;
Final project – 25%;
In-class participation (attending lectures and participating discussion) –5%
A student needs to score on
average at least 60% in total to pass the
class.
Grading scale: A: >93%; A-: >89%; B+:
>85%; B: >81%; B-: >75%; C+: >68%; C: >
60%;
Late
assignments will be marked off 20% for each weekday
that it is late. Each student can ask for the extension of
one assignment up to 5 weekdays.
Please do your
own work. The default consequence for academic dishonesty is a failure for the
course. It is okay to discuss with other students general ideas about
implementing a program. It is NOT okay to copy another
student's program. It is okay to discuss possible program bugs. It is NOT okay to debug another
student's program.
Counseling and Psychological Services (CAPS) can help students who are having difficulties managing stress, adjusting to the demands of a professional program, or feeling sad and hopeless. You can reach CAPS (www.uh.edu/caps) by calling 713-743-5454 during and after business hours for routine appointments or if you or someone you know is in crisis. No appointment is necessary for the “Let's Talk” program, a drop-in consultation service at convenient locations and hours around campus. http://www.uh.edu/caps/outreach/lets_talk.html
Tentative
Schedule
Lectures and Reading
Materials
1. Introduction, visualization pipeline, data types
What is visualization? What does it do? Why is it important? We will have answers for these questions in this week. Also, it is important to tell the difference of scientific data and information data. Can you? Next, we will learn the basic pipeline of visualization (based on my personal opinion and experience). Can you identify the most important step from this pipeline? What are the general types of data and how do people characterize them? What are their representations? Do you know how to parse a given data file?
Lec1_slides (intro)
Lec2_slides (visualization pipeline and data representation)
Additional materials:
We will briefly talk about how we process the visual information. In particular, we will discuss some unique behaviors/characteristics of our visual perception channel which can be used to produce effective visual representation. This topic is new to me, and the materials can be a bit less structured. But you can find many interesting and comprehensive lectures in the additional materials.
Additional materials:
3. Plots and graphs
Plots /graphs are important visualization techniques and forms that people use every day. However, many people ignore some simple principles during plotting, often resulting in sub-optimal graphical representations of their data. We will discuss some of the principles for making effective plots.
Additional materials:
For most of your project assignments, you will need to work on OpenGL library using C/C++. This lecture provides you a short tutorial of OpenGL. Do not expect to fully master it in one or two lectures. You should expect some learning curve. The best way to get yourself familar with this library is to try and fail. There are many examples out there on the Internet. Try to play with them by changing things here and there. In addition, I provide the skeleton code for all the project assignments. Please start playing with it now (not later)!!!
Additional materials:
Many other resources can be found on the Internet, just Google "OpenGL".
WebGL resources:
https://developer.mozilla.org/en-US/docs/Web/API/WebGL_API/Tutorial
How important colors are in visualization? Very important! Without colors (including gray scale colors), there is no visualization! However, not carefully designing the use of the colors will easily result disastrous visualization. In this lecture, we will introduce a number of color spaces and models, along with some principles of proper use of colors in your visual representations.
Additional materials:
6. 2D Scalar Field Visualization
Scalar fields widely exist in many applications. They are the approximation of certain scalar functions that take the (x, y, z) coordinates as the input and return certain quantity values. Visualization of scalar fields plays an important role in various areas. In this lecture, we will specifically focus on scalar fields defined on 2D domain. Two common visualization techniques will be introduced, color plots (i.e. encode scalar information into colors) and iso-contouring (i.e. compute a set of curves, each of which is a collection of all points in the domain that have the same scalar value).
Lec7_slides (color plots and iso-contouring)
Additional materials:
7. 3D Scalar Field Visualization I -- Cut planes and iso-surfacing
To visualize scalar field defined in 3D volume, two standard methods can be used, iso-surfacing and volume rendering. This lecture will focus on the former, which is the extension of the iso-contouring for 2D scalar field visualization -- a geometry-based method. In addition, the lecture will introduce some other interaction for explorating 3D scalar field, such as the cutting plane interaction.
Lec8_slides (cutting planes and iso-surfacing)
8. 3D Scalar Field Visualization II -- direct volume rendering (DVR)
Visualizing 3D volumetric scalar fields as a whole without losing information is sometime necessary, especially for data that contain noise and do not have hard interior boundaries. The technique to do so is a natural extension of the color plots for 2D fields. This technique does not require to create intermediate geometric elements. We, thus, call it "direct" volume rendering. There are a number of computational strategies for DVR, including image-order (e.g., raycasting), object-order (e.g., splatting), and domain-dependent (e.g., shear-warp). For all these methods, the two most important steps are the specification of transfer functions and the composition of colors and opacity.
Lec9_slides (DVR - Raycasting)
Lec10_slides (DVR - Splatting and texture-based)
Lec11_slides (Transfer functions)
Additional materials:
9. Vector field (or flow) visualization
Vector fields are a common form of data that are generated from various applications. They are widely used to study the behaviors of gas and liquids under different circumstances, which is dominated by certain aero- and hydro- dynamical systems. Visualizing and analyzing vector fields is important to the understanding and manipulating of these different dynamical systems. In this lecture, we will learn the basic concepts of vector fields and some simple but effective visualization techniques including the direct method (e.g., arrow plots) and the geometric-based method (e.g., streamline computation and streamline placement).
Additional reading:
10. Texture-based flow visualization (2D)
Unlike the previous topic on scalar field data visualization, vector field data contain both scalar and directional information. To visualize the directional information encoded in the vector field in a coherent and space filling fashion, texture-based method was proposed. In short, texture-based method aligns the orientation of the texture patterns with the direction of the vector field at each pixel of the image. LIC (line integral convolution) is considered the first successful texture-based method. Many variations have been proposed since then. Most of them adopt the LIC with different implementation improvement. IBFV (image-based flow visualization) is another popular technique that utilizes the hardware ability (e.g., the texture memory of the video card). We will mostly focus on these two techniques and their variations in this lecture.
Additional reading:
11. Feature-based vector field visualization
In many situations, the given vector fields (or flow) are complex and with large-scale, so that standard visualization (e.g., direct, geometric-based or texture-based methods) cannot draw the attention of the domain experts to those "features/patterns" that they are interested in. In that case, it is required to first identify (or extract) those features of interest, then highlight them on top of the other visualizations. This leads to the feature-based visualization. There are two types of features in flow: non-topological features and topological features. The non-topological features are typically defined and computed based on some local physical properties of the flow. On the other hand, vector field topology provides the qualitative (or structural) information of the flow data. It also provides a partitioning scheme for spatial domain segmentation so that the flow has homogeneous behavior within a region (i.e., flowing from one repeller/source to one attractor/sink). We will specifically focus on the extraction of topology for 2D steady vector fields for this topic.
Lec14_15_slides (non-topology-based and 2D topology)
Additional reading:
12. 3D vector field visualization
The previous introduced visualization techniques can be applied to both 2D and 3D vector fields. However, 3D flows have much more complex features and structure than their 2D counterpart. In addition, the occlusion issue in 3D visualization makes the clear and intuitive representation of 3D flow behaviors challenging. In this topic, we will briefly review a number of geometric-based methods for the effective visualization of 3D flows.
Additional reading:
13. Unsteady vector field visualization
The world is changing. Thus, time-dependent flows are the dominant type of flows in real-world. The additional time dimension greatly increases the difficulty of their analysis and visualization, because the originally well-defined features in steady flows may not be valid any more under time-varying setting (e.g., fixed points and periodic orbits that require infinitely long time in their characterization). We will briefly review a number of important geometric descriptors (i.e., pathlines, streaklines and timelines) and their computation for the visualization of unsteady flow. Other important techniques for unsteady flow visualization, such as FTLE, the Lagrangian accumulation framework, and Galilean invariant property, may be discussed given the available time.
Additional reading:
14. Tensor field background and its applications
Tensor fields are now a popular subject in the visualization community. It measures the local higher ordered property on a higher dimensional manifold compared to scalar (0D property) and vector (1D property). Examples include stress/strain tensor, curvature tensor, metric tensor, diffusion tensor, and velocity gradient tensor. This lecture will emphasize the importance of tensors with a number of applications, and introduce the basic concepts and computations of this complex data form with the focus on second order tensors.
Additional reading:
15. Tensor field visualization: glyph packing
Glyphs are good at describing the local patterns of tensors. With proper placement technique, the trend of the placed glyphs can reveal certain global patterns of the tensor field. We will look at a few glyph design and packing schemes in this lecture, especially the superquardic glyphs and the particle-system based glyph packing.
Additional reading:
16. Tensor field visualization: texture-based and geometric-based methods
This lecture will cover the texture-based and geometric-based visualization techniques for second order tensor visualization. Two extended texture-based methods will be introduced, including the hyperLIC and extended IBFV. Hyperstreamline computation and placement will be introduced for the geometric-based visualization as well as its application in diffusion tensor imaging.
Additional reading:
Graphs are one of those common data representations that we see from many applications. Any relation information (either directional or non-directional) can be represented by a graph with the graph nodes corresponding to the objects and the edges representing the relations between objects. In many real-world examples, the relations between subjects are overwhelmingly complex, typically in higher-dimensional forms. This requires an intuitive visualization to help people understand the patterns and trends encoded in these graphs. In this lecture, we will briefly describe a number of representative and basic graph visualization techniques. Many more advanced techniques can be derived from them.
Additional reading:
Higher dimensional data visualization and analysis is getting more and more attention these days as many data and the relation encoded in them are in high dimension. Due to our limited knowledge in the higher dimensional space beyond our physical world and the constraints of 2D display devices, producing intuitive visualization for these types of data is specifically challenging. We will review some of issues and research problems of this topic, and introduce one of the standard techniques for visualizing high dimensional data, i.e. the parallel coordinate.
Additional reading:
Assignments (with sample skeleton code and data)
There are a number of programming assignments for this course. The skeleton codes for these assignments can be found below. Note that we will mostly use the PLY format as the format for our data files. To learn more about the PLY format, you can go to here. There are a couple different versions of the skeleton codes with different user interfaces for you to choose. Please choose just one version that you are most comfortable to work with and stick with that interface to minimize your learning curve.
Other options for finishing the programming assignments and useful resources (I will not be able to provide much help here!!):
The final project is a group project. Each group can have at most TWO group members. Please select from ONLY the following provided topics!
Please submit your final project proposal VIA BLACKBORAD by Oct 27 11:59PM to earn 5 points for your final project!!
The structure of the presentation (Up to 6 mins + 2 mins Q&A, a stopwatch will be used!!!): The final project presentation will take place on Nov. 26 in the class.
- Problem definition (especially what is the visualization problem you are addressing)
- Describe your technique (mostly on algorithm and visualization/interface design)
- Results and/or demo (Show your current results. Provide necessary interpretation of your visualization. How do you know you have resolved the problem?)
- Future Work (If your results are half-cooked, what else do you still need to do to make it complete before the deadline? If the results are ready/finalized, what do think you can improve further in the future)
Requirements
of final project submission
You will need to submit your source code, your final project presentation (.pptx or .pdf), and your report in a single .zip file via the blackboard system by December 6!
For the final report, please write it in the IEEE TVCG style (4-8 pages including figures and illustrations). You can find the template of this format in the following link (you can find the downloadable templates on the left of the webpage):
https://www.computer.org/web/tvcg/author
The final report should include the following components:
1. High quality, interactive volume rendering for large-scale data
J. Kruger and R. Westermann, Acceleration Techniques for GPU-based Volume Rendering. IEEE Visualization 2003.
P. Schlegel, M. Makhinya and R. Pajarola. Extinction-Based Shading and Illumination in GPU Volume Ray-Casting. IEEE TVCG Vol. 17(12): pp. 1795 - 1802, 2011.
Daniel Jonsson, Erik Sunden, Anders Ynnerman, and Timo Ropinski. Interactive Volume Rendering with Volumetric Illumination. Eurographics STAR program - 2012.
2. Texture-based visualization of surface and 3D vector fields
Jarke J. van Wijk, Image Based Flow Visualization for Curved Surfaces, IEEE Visualization 2003.
Robert S. Laramee, Bruno Jobard, and Helwig Hauser, Image Space Based Visualization of Unsteady Flow On Surfaces. in Proceedings of IEEE Visualization (IEEE Vis 2003).
3. Geometric-based 3D vector field visualization
- Streamline-based 3D vector field visualization
Yuan Chen, Jonathan Cohen, Julian Krolik, Similarity-Guided Streamline Placement with Error Evaluation, IEEE TVCG Vol. 13(6): pp. 1448-1455.
- Stream surface construction and rendering
Christoph Garth, Hari Krishnan, Xavier Tricoche, T. Bobach, Ken Joy, Generation of Accurate Integral Surfaces in Time-Dependent Vector Fields, in Proceedings of IEEE Visualization '08, 2008.
4. Vector field topology computation and visualization (2D)
5. Unsteady vector field visualization (2D)
- FTLE computation and visualization
Christoph Garth, Florian Gerhardt, Xavier Tricoche, Hans Hagen, Efficient Computation and Visualization of Coherent Structures in Fluid Flow Applications, in "IEEE Transactions on Visualization and Computer Graphics (Proceedings IEEE Visualization 2007)", Volume 13, Number 6, pp 1464--1471, 2007.
- Feature tracking
T. Weinkauf, H. Theisel, A. Van Gelder, and A. Pang. Stable Feature Flow Fields. IEEE Transactions on Visualization and Computer Graphics 17(6), June 2011 .
6. Tensor field visualization
- Glyph-packing
Gordon Kindlmann and Carl-Fredrik Westin. Diffusion tensor visualization with glyph packing, IEEE Visualization 2006.
- Texture-based (for 2D and curved surfaces only)
Xiaoqiang Zheng and Alex Pang, HyperLIC.
- Hyperstreamline-based
G. Reina, K. Bidmon, F. Enders, P. Hastreiter, and T. Ertl. GPU-Based Hyperstreamlines for Diffusion Tensor Imaging. EuroVis 2006.
7. Web-browser-based Visualization Systems (using Javascript and WebGL if showing 3D content)
- High-dimensional information visualization
- 3D scientific data visualization (you can extend your WebGL template from the homework for this topic)
8. Virtual reality /augmented reality in visualization
9. Visualization for different applications
Examples include software engineering, cyber security, etc. You need to find the proper data sets and problems for this topic
10. Machine-learning techniques for data analysis and visualization
Machine-learning techniques have been applied for different analysis and visualization tasks. However, in order to apply ML techniques, your problem has to be defined in a way that the traditional methods cannot solve effectively. For example, you aim to identify certain features or structures in the data that are not well-defined, but experts can identify them manually. Please come and discuss with me if you want to work on this topic.
11. IEEE Visualization contests (bonus will be given to students who work on the following contests)
Visualize the aftermath of volcanic eruptions
Visualize ensemble particle data
Visualize clouds and atmospheric processes
Visualization and analysis of deep water asteroid impacts
Complete list of IEEE SciVis contest
List of visual analytic (VAST) challenges