COURSE PROJECT - CS5187 VISION AND IMAGE
项目要求
1. EXPECTATION
The course project can be in group or individual. Through the project, you are expected to show knowledge (B grade), technical capability (B to A grade) and creativity (A grade) in any topic related to computer vision and/or image processing.
2. GROUPING
Each group can have at most three members. It is OK to work alone. The grades of group members are not necessarily the same depending on contribution. The expectation for the scope of a project, as well as the grade, will be proportional to the group size.
3. SCOPE
No two groups can have the exactly same title and content. All topics need to be approved by Prof. Ngo. If a proposed topic is overlapped with another group, the approval will be based on First-Come-First-Approve. The scope of a project can be in one (or multiple) of the following formats:
Algorithm Implementation and Result Comparison
Implement at least two algorithms/techniques (learn in class or elsewhere) and conduct empirical evaluation for performance comparison.
Proof-of-concept
You can propose a new algorithm/technique and show the improvement in theory and/or by experiment. You may or may not implement the algorithm, but a complete description of proposal, algorithm analysis and expected result is required.
Application of algorithms/techniques for innovative application
You can implement a vision or image related system to demonstrate your innovation. You are required to showcase the system by the end of project.
Critical survey
A survey of at least five research papers for a specific problem of computer vision and/or image processing. The survey must include critical analysis of strength and limitation of different techniques.
4. SCHEDULE
5. SUGGESTED TOPICS
- Image segmentation
- 3D estimation from image sequence
- 3D volume estimation
- Application using dual-camera mobile phone (e.g., using ARKit in iphone)
- Image or video retrieval
- Image or video captioning
- X detection and tracking (X = face, gesture or pose, vehicle, pedestrian)
- X recognition for interactivity (X = face, gesture or pose)
- Facial expression recognition
- X recognition with deep learning (X = object, scene, activity, 3D model)
- Seam carving
- Domain specific image processing (medicine, food, traffic, cultural heritage)
- Person re-identification in multi-camera setting
- Video surveillance
- Image editing with computer vision techniques
- Multi-sensor (one of the sensors must be camera) analysis or application
- Super-resolution image reconstruction
2-15开会：确定题目 Faical Expression Recognition
https://www.kaggle.com/c/challenges-in-representation-learning-facial-expression-recognition-challenge/leaderboard
https://github.com/amineHorseman/facial-expression-recognition-using-cnn#fer2013
https://github.com/search?q=Facial+Expression
2-22开会：
下周安排
每个人在github上找合适的2013kaggle比赛相关的解决方法（自己本地跑通）
确定题目后编辑email并发送，安排每个人的role
3-15开会：
1.确定自己要做哪个项目
2.看懂项目代码逻辑，理解模型(CNN/VGG)的原理
3.写一个文档大致介绍一下
Dongyu Zhao https://github.com/WuJie1010/Facial-Expression-Recognition.Pytorch
ShaoHeng He https://github.com/oarriaga/face_classification
Kejun Du 待定