• Corpus ID: 225062031

Comprehensive Attention Self-Distillation for Weakly-Supervised Object Detection

@article{Huang2020ComprehensiveAS,
  title={Comprehensive Attention Self-Distillation for Weakly-Supervised Object Detection},
  author={Zeyi Huang and Yang Zou and B. V. K. Vijaya Kumar and Dong Huang},
  journal={ArXiv},
  year={2020},
  volume={abs/2010.12023}
}
Weakly Supervised Object Detection (WSOD) has emerged as an effective tool to train object detectors using only the image-level category labels. However, without object-level labels, WSOD detectors are prone to detect bounding boxes on salient objects, clustered objects and discriminative object parts. Moreover, the image-level category labels do not enforce consistent object detection across different transformations of the same images. To address the above issues, we propose a Comprehensive… 

Figures and Tables from this paper

Spatial likelihood voting with self-knowledge distillation for weakly supervised object detection
Scaling Novel Object Detection with Weakly Supervised Detection Transformers
TLDR
The Weakly Supervised Detection Transformer is proposed, which enables efficient knowledge transfer from a large-scale pretraining dataset to WSOD finetuning on hundreds of novel objects.
Discovery-and-Selection: Towards Optimal Multiple Instance Learning for Weakly Supervised Object Detection
TLDR
This paper proposes a discovery-and-selection approach fused with multiple instance learning (DS-MIL), which selects optimal solution from multiple local minima and can consistently improve the baselines, reporting state-of-the-art performance.
LGD: Label-guided Self-distillation for Object Detection
TLDR
Compared with a classical teacher-based method FGFI, LGD not only performs better without requiring pretrained teacher but also reduces 51% training cost beyond inherent student learning.
Boosting Weakly Supervised Object Detection via Learning Bounding Box Adjusters
TLDR
This paper defends the problem setting for improving localization performance by leveraging the bounding box regression knowledge from a well-annotated auxiliary dataset and proposes an EM-like multi-stage training algorithm for LBBA-boosted WSOD.
Contrastive Proposal Extension with Sequential Network for Weakly Supervised Object Detection
TLDR
Inspired by the habit of observing things by the human, a new method by comparing the initial proposals and the extension ones to optimize those initial proposals is proposed, which will guide the basic WSOD to suppress bad proposals and improve the scores of good ones.
Absolute Wrong Makes Better: Boosting Weakly Supervised Object Detection via Negative Deterministic Information
TLDR
This paper proposes a negative deterministic information (NDI) based method for improving WSOD, namely NDI-WSOD, which consists of two stages: NDI collecting and exploiting.
Contrastive Proposal Extension with LSTM Network for Weakly Supervised Object Detection
TLDR
Inspired by the habit of observing things by the human, a new method by comparing the initial proposals and the extension ones to optimize those initial proposals is proposed, which will guide the basic WSOD to suppress bad proposals and improve the scores of good ones.
Mixed Supervised Object Detection by Transferring Mask Prior and Semantic Similarity
TLDR
Mixed supervision object detection with mixed supervision is considered, which learns novel object categories using weak annotations with the help of full annotations of existing base object categories, and further transfer mask prior and semantic similarity to bridge the gap between novel categories and base categories.
Weakly Supervised Rotation-Invariant Aerial Object Detection Network
TLDR
This paper constructs a novel end-to-end weakly supervised Rotation-Invariant aerial object detection Network (RINet), and proposes to couple the predicted instance labels among different rotation-perceptive branches for generating rotation-consistent supervision and meanwhile pursuing all possible instances.
...
...

References

SHOWING 1-10 OF 55 REFERENCES
Attention-Based Dropout Layer for Weakly Supervised Object Localization
  • Junsuk ChoeHyunjung Shim
  • Computer Science
    2019 IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)
  • 2019
TLDR
An Attention-based Dropout Layer (ADL), which utilizes the self-attention mechanism to process the feature maps of the model to improve the accuracy of WSOL, achieving a new state-of-the-art localization accuracy in CUB-200-2011 dataset.
PCL: Proposal Cluster Learning for Weakly Supervised Object Detection
TLDR
This paper first shows that instances can be assigned object or background labels directly based on proposal clusters for instance classifier refinement, and then shows that treating each cluster as a small new bag yields fewer ambiguities than the directly assigning label method.
Instance-Aware, Context-Focused, and Memory-Efficient Weakly Supervised Object Detection
TLDR
This work develops an instance-aware and context-focused unified framework for weakly supervised video object detection that achieves state-of-the-art results on COCO, VOC 2007, and VOC 2012 while devising a memory-efficient sequential batch back-propagation.
Weakly Supervised Object Detection With Segmentation Collaboration
TLDR
This work proposes a novel end-to-end weakly supervised detection approach, where a newly introduced generative adversarial segmentation module interacts with the conventional detection module in a collaborative loop, forming a more comprehensive solution.
C-MIL: Continuation Multiple Instance Learning for Weakly Supervised Object Detection
TLDR
A continuation optimization method is introduced into MIL and thereby creating continuation multiple instance learning (C-MIL), with the intention of alleviating the non-convexity problem in a systematic way.
C-MIDN: Coupled Multiple Instance Detection Network With Segmentation Guidance for Weakly Supervised Object Detection
  • Gao YanB. Liu Dongrui Fan
  • Computer Science
    2019 IEEE/CVF International Conference on Computer Vision (ICCV)
  • 2019
TLDR
A novel Coupled Multiple Instance Detection Network (C-MIDN) is proposed, which uses a pair of MIDNs, which work in a complementary manner with proposal removal and is coupled to obtain tighter bounding boxes and localize multiple objects.
Improving Object Detection with Inverted Attention
TLDR
This paper improves object detectors using a highly efficient and fine-grain mechanism called Inverted Attention (IA), which iteratively inverts attention on feature maps which pushes the detector to discover new discriminative clues and puts more attention on complementary object parts, feature channels and even context.
Continual Universal Object Detection
TLDR
A continual object detector that can learn sequentially from different domains without forgetting is proposed and an adaptive exemplar sampling method is proposed to leverage exemplars from previous tasks for less forgetting effectively.
Boosting Weakly Supervised Object Detection with Progressive Knowledge Transfer
TLDR
An effective knowledge transfer framework to boost the weakly supervised object detection accuracy with the help of an external fully-annotated source dataset, whose categories may not overlap with the target domain, is proposed.
Min-Entropy Latent Model for Weakly Supervised Object Detection
TLDR
A min-entropy latent model (MELM) is proposed for weakly supervised object detection, unified with feature learning and optimized with a recurrent learning algorithm, which progressively transfers the weak supervision to object locations.
...
...