Research map / Detection and segmentation / Weakly supervised detection and segmentation
Label, Verify, Correct: A Simple Few Shot Object Detection Method
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Apply full pseudo-labelling with candidate sourcing, label verification, and box correction | Faster R-CNN, ResNet-50 · MS-COCO novel classes, 30-shot | nAP: 16.6 -> 25.5 | Helped |
| Add candidate sourcing as pseudo-annotations (ablation) | Faster R-CNN, ResNet-50 · MS-COCO novel classes, 30-shot | nAP: +1.8 nAP | Helped |
| Add Label Verification to candidate sourcing (ablation) | Faster R-CNN, ResNet-50 · MS-COCO novel classes, 30-shot | nAP: +3.4 nAP | Helped |
| Add Box Correction after Label Verification (ablation) | Faster R-CNN, ResNet-50 · MS-COCO novel classes, 30-shot | nAP: +3.7 nAP | Helped |
| Use ColorJitter, RandomCrop, and Mosaic augmentations with baseline training (ablation) | Faster R-CNN, ResNet-50 · MS-COCO novel classes, 30-shot | nAP: 13.0 -> 13.8 | Helped |
| Use ColorJitter, RandomCrop, and Mosaic with RPN and RoI finetuning (ablation) | Faster R-CNN, ResNet-50 · MS-COCO novel classes, 30-shot | nAP: 13.0 -> 16.2 | Helped |
| Add Mosaic to ColorJitter, RandomCrop, and RPN/RoI finetuning (ablation) | Faster R-CNN, ResNet-50 · MS-COCO small novel instances, 30-shot | nAPs: 4.3 -> 6.2 | Helped |
| Add RoI activation dropout (ablation) | Faster R-CNN, ResNet-50 with C2 training · MS-COCO novel classes, 30-shot | nAP: +0.4 nAP | Helped |
| Add ignore regions for unverified novel detections during retraining (ablation) | Faster R-CNN, ResNet-50 pseudo-labelling · MS-COCO novel classes, 30-shot | nAP: 21.1 -> 25.5 | Helped |
| Finetune RPN on few-shot novel data (ablation) | Faster R-CNN RPN · MS-COCO novel classes, 30-shot; 100 proposals | nAR IoU=0.5: 49.7 -> 71.0 | Helped |
| Finetune RoI and RPN modules in addition to the classifier (ablation) | Faster R-CNN novel training · MS-COCO novel classes, 30-shot | nAP: 13.0 -> 14.3 | Helped |
| Use Label Verification for person candidates (ablation) | Faster R-CNN pseudo-labelling · MS-COCO person class | nAP: -2.4 nAP | Hurt |
Closest papers by meaning
- Few-Shot Object Detection via Classification Refinement and Distractor Retreatment (CVPR 2021 · similarity 0.92)
- Meta Faster R-CNN: Towards Accurate Few-Shot Object Detection with Attentive Feature Alignment (AAAI 2022 · similarity 0.91)
- Weakly Supervised Few-Shot Object Detection with DETR (AAAI 2024 · similarity 0.91)
- Fine-Grained Prototypes Distillation for Few-Shot Object Detection (AAAI 2024 · similarity 0.90)
- Revisiting Few-Shot Object Detection with Vision-Language Models (NeurIPS 2024 · similarity 0.90)
- Few-Shot Object Detection with Foundation Models (CVPR 2024 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.