Research map / Detection and segmentation / Open-set and few-shot detection
LVIS: A Dataset for Large Vocabulary Instance Segmentation
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 |
|---|---|---|---|
| Increase detections per image from 100 to 300 (ablation) | Mask R-CNN ResNet-50-FPN on LVIS · LVIS val v0.5 | mask AP: 14.8 -> 15.7 | Helped |
| Lower minimum score threshold from 0.050 to 0.001 with 300 detections per image (ablation) | Mask R-CNN ResNet-50-FPN on LVIS · LVIS val v0.5 | mask AP: 15.7 -> 20.8 | Helped |
| Use repeat factor sampling with t=0.001 instead of no resampling (ablation) | Mask R-CNN ResNet-50-FPN · LVIS val v0.5 | rare-category AP: 3.2 -> 13.4 | Helped |
| Use repeat factor sampling with t=0.001 instead of no resampling (ablation) | Mask R-CNN ResNet-50-FPN · LVIS val v0.5 | mask AP: 21.0 -> 23.2 | Helped |
| Use repeat factor sampling with t=0.001 instead of no resampling (ablation) | Mask R-CNN ResNet-50-FPN · LVIS val v0.5 | frequent-category AP: 27.7 -> 27.1 | Hurt |
| Use class aware sampling instead of no resampling (ablation) | Mask R-CNN ResNet-50-FPN · LVIS val v0.5 | rare-category AP: 3.2 -> 8.5 | Helped |
| Use class aware sampling instead of no resampling (ablation) | Mask R-CNN ResNet-50-FPN · LVIS val v0.5 | mask AP: 21.0 -> 18.7 | Hurt |
| Add scale jitter (ablation) | Mask R-CNN ResNet-50-FPN with repeat factor sampling t=0.001 · LVIS val v0.5 | mask AP: 23.2 -> 24.4 | Helped |
| Replace ResNet-50-FPN with ResNet-101-FPN (ablation) | Mask R-CNN with scale jitter and repeat factor sampling t=0.001 · LVIS val v0.5 | mask AP: 24.4 -> 26.0 | Helped |
| Replace ResNet-101-FPN with ResNeXt-101-32×8d (ablation) | Mask R-CNN with scale jitter and repeat factor sampling t=0.001 · LVIS val v0.5 | mask AP: 26.0 -> 27.1 | Helped |
| Increase evaluation subset size from 5k to 20k images (ablation) | Fixed Mask R-CNN ResNet-50-FPN · LVIS test v0.5 | mask AP: 24.8±0.51 -> 18.4 | Hurt |
| Train using 1k images instead of the full COCO train2017 set (ablation) | Mask R-CNN · COCO val2017 | mask AP: 36.4% -> 9.8% | Hurt |
Closest papers by meaning
- Mask-Free OVIS: Open-Vocabulary Instance Segmentation Without Manual Mask Annotations (CVPR 2023 · similarity 0.88)
- OpenVIS: Open-vocabulary Video Instance Segmentation (AAAI 2025 · similarity 0.87)
- The Devil is in the Object Boundary: Towards Annotation-free Instance Segmentation using Foundation Models (ICLR 2024 · similarity 0.87)
- DitHub: A Modular Framework for Incremental Open-Vocabulary Object Detection (NeurIPS 2025 · similarity 0.87)
- Mask-Adapter: The Devil is in the Masks for Open-Vocabulary Segmentation (CVPR 2025 · similarity 0.86)
- How to Evaluate the Generalization of Detection? A Benchmark for Comprehensive Open-Vocabulary Detection (AAAI 2024 · similarity 0.86)
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.