CSE291-J00: Deep Learning Lab (Computer Vision)

Long-tailed Detectection

Papers to compare

• Focal Loss for Dense Object Detection
• Equalization Loss for Long-Tailed Object Recognition (LVIS 2019 champion)
• Decoupling Representation and Classifier for Long-Tailed Recognition (LVIS 2020 baseline)
• Long-Tailed Classification by Keeping the Good and Removing the Bad Momentum Causal Effect
• Learning to Segment the Tail
• 1st Place Solution of LVIS Challenge 2020: A Good Box is not a Guarantee of a Good Mask

At least 5 papers should be compared. You can choose one of colored papers.

Reference

• Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
• Mask R-CNN
• Feature Pyramid Networks for Object Detection
• LVIS: A Dataset for Large Vocabulary Instance Segmentation

Instance Segmentation

Papers to compare

• SGPN: Similarity Group Proposal Network for 3D Point Cloud Instance Segmentation
• GSPN: Generative Shape Proposal Network for 3D Instance Segmentation in Point Cloud
• Associatively Segmenting Instances and Semantics in Point Clouds
• 3D-MPA: Multi Proposal Aggregation for 3D Semantic Instance Segmentation
• PointGroup: Dual-Set Point Grouping for 3D Instance Segmentation

Reference

• PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space
• 3D Semantic Segmentation with Submanifold Sparse Convolutional Networks
• Deep Hough Voting for 3D Object Detection in Point Clouds
• H3DNet: 3D Object Detection Using Hybrid Geometric Primitives
• Generative Sparse Detection Networks for 3D Single-shot Object Detection

Part Segmentation

Papers to compare

• PartNet: A Large-scale Benchmark for Fine-grained and Hierarchical Part-level 3D Object Understanding (baseline)
• Learning object bounding boxes for 3d instance segmentation on point clouds
• SGPN: Similarity Group Proposal Network for 3D Point Cloud Instance Segmentation
• GSPN: Generative Shape Proposal Network for 3D Instance Segmentation in Point Cloud
• Point Cloud Instance Segmentation using Probabilistic Embeddings

Reference

The same as Instance Segmentation.

Tracking

Papers to compare

• Scalability in Perception for Autonomous Driving: Waymo Open Dataset
• 3D Multi-Object Tracking: A Baseline and New Evaluation Metrics
• 1st Place Solutions for Waymo Open Dataset Challenges -- 2D and 3D Tracking (1st place)
• PV-RCNN: Point-Voxel Feature Set Abstraction for 3D Object Detection (2nd place)

Reference

• Multiple Object Tracking: A Literature Review
• Simple Online and Realtime Tracking
• Simple Online and Realtime Tracking with a Deep Association Metric