Interests
- Computer Vision
- Visual Recognition
- Representation Learning
Latest
- End-to-end 3D Tracking with Decoupled Queries
- FB-BEV: BEV Representation from Forward-Backward View Transformations
- FocalFormer3D: Focusing on Hard Instance for 3D Object Detection
- Fully Attentional Networks with Self-emerging Token Labeling
- Real-Time Radiance Fields for Single-Image Portrait View Synthesis
- FB-OCC: 3D Occupancy Prediction based on Forward-Backward View Transformation
- Vision Transformers Are Good Mask Auto-Labelers
- VoxFormer: Sparse voxel transformer for camera-based 3D semantic scene completion
- Prismer: A Vision-Language Model with An Ensemble of Experts
- Re-ViLM: Retrieval-Augmented Visual Language Model for Zero and Few-Shot Image Captioning
- MinVIS: A minimal video instance segmentation framework without video-based training
- Test-time prompt tuning for zero-shot generalization in vision-language models
- Partial Convolution for Padding, Inpainting, and Image Synthesis
- Understanding The Robustness in Vision Transformers
- CoordGAN: Self-Supervised Dense Correspondences Emerge from GANs
- FreeSOLO: Learning to Segment Objects without Annotations
- Panoptic SegFormer: Delving deeper into panoptic segmentation with transformers
- M$^2$BEV: Multi-camera joint 3D detection and segmentation with unified birds-eye view representation
- Learning contrastive representation for semantic correspondence
- Coupled Segmentation and Edge Learning Using Dynamic Graph Propagation
- Coupled Segmentation and Edge Learning via Dynamic Graph Propagation
- SegFormer: Simple and efficient design for semantic segmentation with transformers
- Domain Stylization: A Fast Covariance Matching Framework towards Domain Adaptation
- DiscoBox: Weakly Supervised Instance Segmentation and Semantic Correspondence from Box Supervision
- Joint Disentangling and Adaptation for Cross-Domain Person Re-Identification
- UFO2: A Unified Framework towards Omni-supervised Object Detection
- Angular Visual Hardness
- Transposer: Universal Texture Synthesis Using Feature Maps as Transposed Convolution Filter
- Instance-aware, Context-focused, and Memory-efficient Weakly-Supervised Object Detection