Computer Vision | Machine Learning | Computer Graphics | Interdisciplinary Applications

Computer Vision

1. Massive Fine-grained Object Classification
SKU detection and classification in retail scenarios

- Precise Detection in Densely Packed Scenes
- Fine-grained object classification
- Appearance and volume confusion
- [Retail50K Dataset]
2. Vision-based Driver Monitoring
Behaviour recognition
- Smoking in the car
- Calling
- Emotion (e.g. happy, sad, angry, etc.)

Drowsiness detection
- FatigueTree
- FatigueView Dataset
- Eye-blinking, yawning, nodding off, etc.
3. Image stitching and object deduping
Image Stitching
- Interactive image stitching in APP
- Robust image stitching in service

Remove duplicated objects from successive images
- Stitching-based deduping
- Graph-based deduping
4. Camera Pose Estimation
Camera Pose Estimation
[Sensors 2020]

- A convolutional network for feature learning
- ARToolKits
- Objects in industrial environments
- Regress the 6-DOF camera pose
5. Shape-based Object Recognition
Object Recognition using Coarse-grained Shape Features

- Shape Retrieval by Contour Segment Matching
- Shap Classification with Supervised Optimisation
- Gradient Hill Climbing with Simulated Annealing
- 86.3% Classification Rate on the MPEG-7 Dataset
- ICIP 2014, ICPRAM 2015
Object Recognition using Fine-Grained Shape Features
[Project 1] [Project 2]

- Shape-based Object Matching Using Point Context
- Object Matching with Hierarchical Skeletons
- Shape Interesting Point Detection with Overall Shape Trend
- Interesting Point Matching with High Order Graphs
- 99.21% Bulls-Eye Score on the MPEG7 Shape Dataset
- CORES 2013, ICMR 2015
Object Recognition using Integrated Shape Features

- Shape Retrieval by Contour Segment and Skeleton Matching
- Shape Matching Using Point Context and Contour Segments
- ACCV 2014, ICIP 2014
Vision-based Monitoring System for Accurate Vojta-Therapy

- Microsoft Kinect
- Movement Recognition
- Coarse-Grained Shape Features
- ICIS 2016

6. Environmental Microorganisms Classification
Shape-Based Classification of Environmental Microorganisms

- Coarse-Grained Shape Features
- Multi-Class Support Vector Machine

7. Contour Segment Descriptors
Evaluating Contour Segment Descriptors
[short] [full]

- Evaluating 27 Different Contour Segment Descriptors
- Open Curve Matching
- Sketch Matching
- Hungarian, DP and DTW Matching Algorithms
Stripe-based Object Matching

- Stripe-based Object Representation
- without any 3D Object Reconstruction Process
- Robust Point Context Descriptor
- Bhattacharyya Distance
- ICIS 2016

8. Computer Vision between Humans and Machines
Object Similarity by Humans and Machines

- 3D Object Matching and Retrieval
- AAAI MCP 2014
Skeleton Pruning by Human Perception

- Skeleton Pruning by Humans and Machines
- Skeleton Graph Matching
- ICIP 2016