Bowen Dong

Recent Projects

Learnable Bounding Box Adjusters

Project Page | PyTorch Implementation

We aim to build a transfer learning-based weakly-supervised object detection (WSOD) method, to stably improve the localization performance of current WSOD framework, thus imporving the overall detection accuracy. This bounding box adjuster can be easily adopted into various WSOD training frameworks.

Few-shot Vision Transformer

Project Page | PyTorch Implementation

The project starts with the observation that naïve application of ViT to few-shot learning results in substantial performance degradation. To resolve this issue, this project introduces a new few-shot training framework for ViTs, named self-promoted supervision (SUN). This framework aims to enhance the capability of learning the token dependencies via location-specific supervision. As a result, SUN reports high performance on few-shot learning benchmarks. It can be considered as a good baseline for ViTs on few-shot learning.

Research

I'm interested in image and video understanding with non-ideal supervision (e.g., weakly-supervised object detection and few-shot classification).

1. SALAD-Bench: A Hierarchical and Comprehensive Safety Benchmark for Large Language Models
   Lijun Li*, Bowen Dong*, Ruohui Wang*, Xuhao Hu*, Wangmeng Zuo, Dahua Lin, Yu Qiao, Jing Shao
   ACL Findings 2024 | paper | code

2. ConSept: Continual Semantic Segmentation via Adapter-based Vision Transformer
   Bowen Dong, Guanglei Yang, Wangmeng Zuo, Lei Zhang
   submitted to TIP 2024 | paper | code

3. TextField3D: Towards Enhancing Open-Vocabulary 3D Generation with Noisy Text Fields
   Tianyu Huang, Yihan Zeng, Bowen Dong, Hang Xu, Songcen Xu, Rynson W.H. Lau, Wangmeng Zuo
   ICLR 2024 | paper

4. Texts as Images in Prompt Tuning for Multi-Label Recognition
   Zixian Guo, Bowen Dong, Zhilong Ji, Jinfeng Bai, Yiwen Guo, Wangmeng Zuo
   CVPR 2023 | paper | code

5. LPT: Long-tailed Prompt Tuning for Image Classification
   Bowen Dong, Pan Zhou, Shuicheng Yan, Wangmeng Zuo
   ICLR 2023 | paper | code

6. CLIP2Point: Transfer CLIP to Point Cloud Classification with Image-Depth Pre-training
   Tianyu Huang, Bowen Dong, Yunhan Yang, Xiaoshui Huang, Rynson W.H. Lau, Wanli Ouyang, Wangmeng Zuo
   ICCV 2023 | paper | code

7. Symmetry-Aware Transformer-based Mirror Detection
   Tianyu Huang, Bowen Dong, Jiaying Lin, Xiaohui Liu, Rynson Lau, Wangmeng Zuo
   AAAI 2023 | paper | code

8. Self-Promoted Supervision for Few-Shot Transformer
   Bowen Dong, Pan Zhou, Shuicheng Yan, Wangmeng Zuo
   ECCV 2022 | paper | code

9. W2N: Switching From Weak Supervision to Noisy Supervision for Object Detection
   Zitong Huang, Yiping Bao, Bowen Dong, Erjin Zhou, Wangmeng Zuo
   ECCV 2022 | paper | code

10. Boosting Weakly Supervised Object Detection via Learning Bounding Box Adjusters
    Bowen Dong, Zitong Huang, Yuelin Guo, Qilong Wang, Zhenxing Niu, Wangmeng Zuo
    ICCV 2021 | paper | code

Awards

2018, Silver Prize, Award on LanQiao Cup Final

2017, Bronze Prize, Award on ACM-ICPC Shenyang

2017, Bronze Prize, Award on ACM-CCPC Hangzhou

2017, Silver Prize, Award on ACM-CCPC Heilongjiang

2017, Silver Prize, Award on ACM-CCPC Northeast Region