hust-lidelong / WSOD_Paper_List

A list of recent WSOD paper

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

WSOD Paper List

A list of Weakly Supervised Object Detection (WSOD) papers published in recent years.
With title, pdf link, code link and performance.

Hint

  1. Only show single model performance in VOC2007.
  2. For any questions, feel free to create an issue or contact me by email.

Todo

  • Simple summary of some of these papers.
  • Some other weakly supervised vision understanding tasks, such as localization, segmentations.

Catalogs

Image label

2020

  • Comprehensive Attention Self-Distillationfor Weakly-Supervised Object Detection

  • Enabling Deep Residual Networks for Weakly Supervised Object Detection

  • Instance-aware, Context-focused, and Memory-efficient Weakly Supervised Object Detection

    • CVPR 2020 [pdf] [code_pytorch]
    • Performance: 54.9(MAP) 68.8(CorLoc)
      1. MIST = OICR (top p% boxes) + bbox regress
      1. Concrete DropBlock, learnable drop block
      1. Sequential batch bp
  • SLV: Spatial Likelihood Voting for Weakly Supervised Object Detection

    • CVPR 2020 [pdf]
    • Performance: 53.9(MAP) 71.0(CorLoc)
  • Distilling Knowledge from Refinement in Multiple Instance Detection Networks

    • CVPR 2020 Workshop [pdf] [code pytorch]
    • Performance: 49.7(MAP) 65.7(CorLoc)

2019

  • Object Instance Mining for Weakly Supervised Object Detection

    • AAAI 2019 [pdf] [code_caffe]
    • Performance: 50.1(MAP) 67.2(CorLoc)
  • Towards Precise End-to-end Weakly Supervised Object Detection Network

    • ICCV 2019 [pdf]
    • Performance: 51.5(MAP) 68.0(CorLoc)
  • WSOD2: Learning Bottom-up and Top-down Objectness Distillation for Weakly-supervised Object Detection

    • ICCV 2019 [pdf]
    • Performance: 53.6(MAP) 69.5(CorLoc)
    • use objectness information to guide bbox regress
  • Object-Aware Instance Labeling for Weakly Supervised Object Detection

    • ICCV 2019 [pdf]
    • Performance: 47.6(MAP) 66.7(CorLoc)
  • SDCN: Weakly Supervised Object Detection with Segmentation Collaboration

    • ICCV 2019 [pdf]
    • Performance: 50.2(MAP) 68.6(CorLoc)
  • C-MIDN: Coupled Multiple Instance Detection NetworkWith Segmentation Guidance forWeakly Supervised Object Detection

    • ICCV 2019 [pdf]
    • Performance: 52.6(MAP) 68.7(CorLoc)
  • C-MIL: Continuation Multiple Instance Learning for Weakly Supervised Object Detection

    • CVPR 2019 [pdf] [code_torch]
    • Performance: 50.5(MAP) 65.0(CorLoc)
    • use continuation optimization to replace detection stream at WSDDN
  • Dissimilarity Coefficient based Weakly Supervised Object Detection

    • CVPR 2019 [pdf]
    • Performance: 52.9(MAP) 70.9(CorLoc)
  • You reap what you sow: Using Videos to Generate High Precision Object Proposals for Weakly-supervised Object Detection

    • CVPR 2019 [pdf] [code]
    • Performance: 46.9(MAP) 66.5(CorLoc)
  • MELM: Min-Entropy Latent Model for Weakly Supervised Object Detection

  • Utilizing the Instability in Weakly Supervised Object Detection

    • CVPR 2019 Workshop [pdf]
    • Performance: 52.0(MAP) 66.9(CorLoc)

2018

  • TS2C: Tight Box Mining with Surrounding Segmentation Context for Weakly Supervised Object Detection

    • ECCV2018 [pdf]
    • Performance: 44.3(MAP) 61.0(CorLoc)
  • WSRPN: Weakly Supervised Region Proposal Network and Object Detection

    • ECCV2018 [pdf]
    • Performance: 47.9(MAP) 66.9(CorLoc)
  • W2F: A Weakly-Supervised to Fully-Supervised Framework for Object Detection

    • CVPR2018 [pdf]
    • Performance: 52.4(MAP) 70.3(CorLoc)
  • Zigzag Learning for Weakly Supervised Object Detection

    • CVPR2018 [pdf]
    • Performance: 47.6(MAP) 61.2(CorLoc)
  • PCL: Proposal Cluster Learning for Weakly Supervised Object Detection

2017

  • OICR: Multiple Instance Detection Network with Online Instance Classifier Refinement
    • CVPR 2017 [pdf] [code_caffe]
    • Performance: 41.2(MAP) 60.6(CorLoc)
    • Online refinement, a kind of teacher student self-learning

2016

  • ContextLocNet: Context-Aware Deep Network Models for Weakly Supervised Localization

    • CVPR 2016 [pdf] [code_caffe]
    • Performance: 36.3(MAP) 55.1(CorLoc)
  • WSDDN: Weakly Supervised Deep Detection Networks

Object Count Label

  • C-WSL: Count-guided Weakly Supervised Localization
    • ECCV 2018 [pdf]
    • Performance: 47.9(MAP) 66.9(CorLoc)

Low-shot Object Detection

  • Many-shot from Low-shot: Learning to Annotate using Mixed Supervision for Object Detection

    • ECCV 2020 [pdf]
    • Performance: 63.3(MAP, 10%) 59.7(MAP, 10-shot)
  • Low Shot Box Correction for Weakly Supervised Object Detection

    • IJCAI 2019 [pdf] [code_pytorch]
    • Performance: 61.8(MAP, 10%) 57.1(MAP, 10-shot)

WSOD with Domain Transfer

  • Boosting Weakly Supervised Object Detection with Progressive Knowledge Transfer

About

A list of recent WSOD paper