Corpus ID: 212725414

Frustratingly Simple Few-Shot Object Detection

@article{Wang2020FrustratinglySF,
  title={Frustratingly Simple Few-Shot Object Detection},
  author={Xin Wang and Thomas E. Huang and Trevor Darrell and Joseph Gonzalez and Fisher Yu},
  journal={ArXiv},
  year={2020},
  volume={abs/2003.06957}
}
Detecting rare objects from a few examples is an emerging problem. Prior works show meta-learning is a promising approach. But, fine-tuning techniques have drawn scant attention. We find that fine-tuning only the last layer of existing detectors on rare classes is crucial to the few-shot object detection task. Such a simple approach outperforms the meta-learning methods by roughly 2~20 points on current benchmarks and sometimes even doubles the accuracy of the prior methods. However, the high… Expand
Generalized Few-Shot Object Detection without Forgetting
TLDR
The approach has demonstrated that the long desired neverforgetting learner is available in object detection and significantly outperforms state-ofthe-art methods on overall performance among all settings as it can achieve competitive results on few-shot classes and does not degrade the base class performance at all. Expand
Hallucination Improves Few-Shot Object Detection
TLDR
This work introduces a hallucinator network that learns to generate additional, useful training examples in the region of interest (RoI) feature space, and incorporates it into a modern object detection model by transferring the shared within-class variation from base classes. Expand
Class-Incremental Few-Shot Object Detection
TLDR
This paper presents the transfer strategy to reduce unnecessary weight adaptation and improve the transfer capability for iFSD and integrates the knowledge distillation technique using a less resource-consuming approach to alleviate forgetting and proposes a novel clustering-based exemplar selection process to preserve more discriminative features previously learned. Expand
Transformation Invariant Few-Shot Object Detection
  • Aoxue Li, Zhenguo Li
  • Computer Science
  • CVPR
  • 2021
TLDR
This work proposes a simple yet effective Transformation Invariant Principle (TIP) that can be flexibly applied to various meta-learning models for boosting the detection performance on novel class objects. Expand
Few-shot Weakly-Supervised Object Detection via Directional Statistics
TLDR
This work proposes a probabilistic multiple instance learning approach for few-shot Common Object Localization (COL) and fewshot Weakly Supervised Object Detection (WSOD) and finds that operating on features extracted from the last layer of a pre-trained Faster-RCNN is more effective compared to previous episodic learning based few- shot COL methods. Expand
Closing the Generalization Gap in One-Shot Object Detection
TLDR
The results show that the models switch from memorizing individual categories to learning object similarity over the category distribution, enabling strong generalization at test time, and suggest that the key to strong few-shot detection models may not lie in sophisticated metric learning approaches, but instead in scaling the number of categories. Expand
Cooperating RPN's Improve Few-Shot Object Detection
TLDR
This work uses multiple distinct yet cooperating RPN's trained to be different, but not too different; doing so yields significant performance improvements over state of the art for COCO and PASCAL VOC in the very few-shot setting. Expand
GenDet: Meta Learning to Generate Detectors From Few Shots.
TLDR
GenDet is proposed, a new meta-learning-based framework that can effectively generate object detectors for novel classes from few shots and, thus, conducts few-shot detection tasks explicitly. Expand
Multi-Scale Positive Sample Refinement for Few-Shot Object Detection
TLDR
This work proposes a Multi-scale Positive Sample Refinement (MPSR) approach to enrich object scales in FSOD, which generates multi-scale positive samples as object pyramids and refines the prediction at various scales. Expand
Meta-DETR: Few-Shot Object Detection via Unified Image-Level Meta-Learning
TLDR
A novel meta-detector framework, namely Meta-DETR, which eliminates region-wise prediction and instead meta-learns object localization and classification at image level in a unified and complementary manner and design a simple but effective Semantic Alignment Mechanism to facilitate meta-learning with deep networks. Expand
...
1
2
3
4
5
...

References

SHOWING 1-10 OF 28 REFERENCES
Meta-Learning to Detect Rare Objects
TLDR
A conceptually simple but powerful meta-learning based framework that simultaneously tackles few- shot classification and few-shot localization in a unified, coherent way and introduces a weight prediction meta-model that enables predicting the parameters of category-specific components from few examples. Expand
Few-Shot Object Detection via Feature Reweighting
TLDR
This work develops a few-shot object detector that can learn to detect novel objects from only a few annotated examples, using a meta feature learner and a reweighting module within a one-stage detection architecture. Expand
A Closer Look at Few-shot Classification
TLDR
The results reveal that reducing intra-class variation is an important factor when the feature backbone is shallow, but not as critical when using deeper backbones, and a baseline method with a standard fine-tuning practice compares favorably against other state-of-the-art few-shot learning algorithms. Expand
LVIS: A Dataset for Large Vocabulary Instance Segmentation
TLDR
This work introduces LVIS (pronounced ‘el-vis’): a new dataset for Large Vocabulary Instance Segmentation, which has a long tail of categories with few training samples due to the Zipfian distribution of categories in natural images. Expand
Prototypical Networks for Few-shot Learning
TLDR
This work proposes Prototypical Networks for few-shot classification, and provides an analysis showing that some simple design decisions can yield substantial improvements over recent approaches involving complicated architectural choices and meta-learning. Expand
Meta R-CNN: Towards General Solver for Instance-Level Low-Shot Learning
TLDR
This work extends Faster /Mask R-CNN by proposing meta-learning over RoI (Region-of-Interest) features instead of a full image feature, and introduces a Predictor-head Remodeling Network (PRN) that shares its main backbone with Faster / Mask R- CNN. Expand
Dynamic Few-Shot Visual Learning Without Forgetting
TLDR
This work proposes to extend an object recognition system with an attention based few-shot classification weight generator, and to redesign the classifier of a ConvNet model as the cosine similarity function between feature representations and classification weight vectors. Expand
Matching Networks for One Shot Learning
TLDR
This work employs ideas from metric learning based on deep neural features and from recent advances that augment neural networks with external memories to learn a network that maps a small labelled support set and an unlabelled example to its label, obviating the need for fine-tuning to adapt to new class types. Expand
Siamese Neural Networks for One-Shot Image Recognition
TLDR
A method for learning siamese neural networks which employ a unique structure to naturally rank similarity between inputs and is able to achieve strong results which exceed those of other deep learning models with near state-of-the-art performance on one-shot classification tasks. Expand
Low-Shot Visual Recognition by Shrinking and Hallucinating Features
TLDR
This work presents a low-shot learning benchmark on complex images that mimics challenges faced by recognition systems in the wild, and proposes representation regularization techniques and techniques to hallucinate additional training examples for data-starved classes. Expand
...
1
2
3
...