LSTD: A Low-Shot Transfer Detector for Object Detection

@article{Chen2018LSTDAL,
  title={LSTD: A Low-Shot Transfer Detector for Object Detection},
  author={Hao Chen and Yali Wang and Guoyou Wang and Yu Qiao},
  journal={ArXiv},
  year={2018},
  volume={abs/1803.01529}
}
Recent advances in object detection are mainly driven by deep learning with large-scale detection benchmarks. [] Key Method First, we design a flexible deep architecture of LSTD to alleviate transfer difficulties in low-shot detection. This architecture can integrate the advantages of both SSD and Faster RCNN in a unified deep framework.

Figures and Tables from this paper

Progressive Object Transfer Detection
TLDR
This work proposes a novel Progressive Object Transfer Detection framework, which outperforms the recent state-of-the-art approaches in object detection and exploits the reliable object supervision from LSTD, which can further enhance the robustness of target detector in the WSTD stage.
A Survey of Deep Learning for Low-Shot Object Detection
TLDR
This survey provides a comprehensive review of Low-Shot Object Detection methods and proposes a thorough taxonomy of LSOD methods and analyzes them systematically, comprising some extensional topics of LS OD (semi-supervised LSOD, weakly-super supervised LSOD and incremental LSOD).
Network Parameter Generation for One-Shot Object Detection
TLDR
A network parameter generation framework for effective one-shot object detection by leveraging the semantic relationship between the source-domain and the target-domain (new classes), which can generate network parameters for the new classes directly.
Low-shot Object Detection via Classification Refinement
TLDR
Experiments show the proposed low-shot classification correction network (LSCN) can bring significant performance gains to rarely labeled categories and outperforms previous work on COCO and PASCAL VOC by a large margin.
Comparison Network for One-Shot Conditional Object Detection
TLDR
A novel one-shot conditional object detection framework, referred as Comparison Network (ComparisonNet), has been proposed, which can detect objects of both seen and unseen classes without further training and has the advantages including class-agnostic, training-free for unseen classes, and without catastrophic forgetting.
Meta-SSD: Towards Fast Adaptation for Few-Shot Object Detection With Meta-Learning
TLDR
The proposed FSD framework employs Single-Shot MultiBox Detector (SSD) as the object detector in this paper, thus called Meta-SSD, and a novel benchmark is constructed from Pascal VOC dataset for training and evaluation of meta-learning FSD.
DeFRCN: Decoupled Faster R-CNN for Few-Shot Object Detection
TLDR
Decoupled Faster R-CNN (DeFRCN) is proposed, which is remarkably superior to other existing approaches and establishes a new state-of-the-art in few-shot literature.
Restoring Negative Information in Few-Shot Object Detection
TLDR
This paper restores the negative information in few-shot object detection by introducing a new negative- and positive-representative based metric learning framework and a new inference scheme with negative and positive representatives.
Borrow from Source Models: Efficient Infrared Object Detection with Limited Examples
TLDR
A transfer approach, Source Model Guidance (SMG), where a high-capacity RGB detection model is leverage as the guidance to supervise the training process of an infrared detection network and the foreground soft label generated from the RGB model is introduced as source knowledge to provide guidance for cross-domain transfer.
...
...

References

SHOWING 1-10 OF 38 REFERENCES
Weakly Supervised Object Localization with Progressive Domain Adaptation
TLDR
This paper addresses the problem of weakly supervised object localization where only image-level annotations are available for training by progressive domain adaptation with two main steps: classification adaptation and detection adaptation.
LSDA: Large Scale Detection through Adaptation
TLDR
This paper proposes Large Scale Detection through Adaptation (LSDA), an algorithm which learns the difference between the two tasks and transfers this knowledge to classifiers for categories without bounding box annotated data, turning them into detectors.
Rich Feature Hierarchies for Accurate Object Detection and Semantic Segmentation
TLDR
This paper proposes a simple and scalable detection algorithm that improves mean average precision (mAP) by more than 30% relative to the previous best result on VOC 2012 -- achieving a mAP of 53.3%.
Few-shot Object Detection
TLDR
This paper studies object detection using a large pool of unlabeled images and only a few labeled images per category, named “few-shot object detection”, and embeds multiple detection models in this framework, which has proven to outperform the single model baseline and the model ensemble method.
Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks
TLDR
This work introduces a Region Proposal Network (RPN) that shares full-image convolutional features with the detection network, thus enabling nearly cost-free region proposals and further merge RPN and Fast R-CNN into a single network by sharing their convolutionAL features.
Large Scale Semi-Supervised Object Detection Using Visual and Semantic Knowledge Transfer
TLDR
Strong evidence is found that visual similarity and semantic relatedness are complementary for the task, and when combined notably improve detection, achieving state-of-the-art detection performance in a semi-supervised setting.
SSD: Single Shot MultiBox Detector
TLDR
The approach, named SSD, discretizes the output space of bounding boxes into a set of default boxes over different aspect ratios and scales per feature map location, which makes SSD easy to train and straightforward to integrate into systems that require a detection component.
Few-Example Object Detection with Model Communication
TLDR
This paper studies object detection using a large pool of unlabeled images and only a few labeled images per category, named “few-example object detection”, and embeds multiple detection models in this framework, which has proven to outperform the single model baseline and the model ensemble method.
Weakly Supervised Object Localization with Latent Category Learning
TLDR
The latent category learning (LCL), which is an unsupervised learning problem given only image-level class labels, is proposed, which uses the typical probabilistic Latent Semantic Analysis to learn the latent categories, which can represent objects, object parts or backgrounds.
Weakly Supervised Deep Detection Networks
  • Hakan BilenA. Vedaldi
  • Computer Science
    2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
  • 2016
TLDR
This paper proposes a weakly supervised deep detection architecture that modifies one such network to operate at the level of image regions, performing simultaneously region selection and classification.
...
...