Research map / Detection and segmentation / Tracking
Object Detection in Videos With Tubelet Proposal Networks
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 |
|---|---|---|---|
| Add temporal classification with an encoder-decoder LSTM instead of a vanilla LSTM (ablation) | MoveTubelets · ImageNet VID validation | Mean AP: 0.678 -> 0.684 | Helped |
| Use motion-regressed tubelets instead of ground-truth-location tubelets (ablation) | LSTM tubelet classifier · ImageNet VID validation | Mean AP: 0.653 -> 0.678 | Helped |
| Add LSTM temporal classification to TPN tubelets (ablation) | MoveTubelets · ImageNet VID validation | Mean AP: 0.623 -> 0.678 | Helped |
| Add TPN tubelets in place of frame-wise static proposals (ablation) | Fast R-CNN detector · ImageNet VID validation | Mean AP: 0.630 -> 0.623 | Hurt |
| Use block initialization rather than random initialization for the 5-frame regression layer (ablation) | MoveTubelets, window 5 · ImageNet VID validation tubelet proposals | Mean IOU: 0.6972 -> 0.8244 | Helped |
| Use block-initialized 5-frame regression rather than the RNN regression baseline (ablation) | MoveTubelets · ImageNet VID validation tubelet proposals | Mean IOU: 0.8060 -> 0.8244 | Helped |
| Increase the temporal window from 5 to 11 frames (ablation) | MoveTubelets with block initialization · ImageNet VID validation tubelet proposals | Mean IOU: 0.8244 -> 0.8017 | Hurt |
| Increase the temporal window from 5 to 20 frames (ablation) | MoveTubelets with block initialization · ImageNet VID validation tubelet proposals | Mean IOU: 0.8244 -> 0.7731 | Hurt |
| Use a longer temporal context with block initialization rather than the 2-frame random model (ablation) | MoveTubelets · ImageNet VID validation tubelet proposals | Mean IOU: 0.7966 -> 0.8244 | Helped |
| Use the proposed detector rather than the strongest listed prior method | Video object localization · YouTubeObjects | CorLoc Avg.: 76.8 -> 92.4 | Helped |
| Use TPN tubelets rather than KCF-tracker tubelets | Fast R-CNN detector · ImageNet VID validation | Mean AP: 0.567 -> 0.623 | Helped |
Closest papers by meaning
- Object Detection From Video Tubelets With Convolutional Neural Networks (CVPR 2016 · similarity 0.89)
- TubeR: Tubelet Transformer for Video Action Detection (CVPR 2022 · similarity 0.88)
- Mobile Video Object Detection With Temporally-Aware Feature Maps (CVPR 2018 · similarity 0.86)
- TubeTK: Adopting Tubes to Track Multi-Object in a One-Step Training Model (CVPR 2020 · similarity 0.86)
- Context Enhanced Transformer for Single Image Object Detection in Video Data (AAAI 2024 · similarity 0.86)
- TGBFormer: Transformer-GraphFormer Blender Network for Video Object Detection (AAAI 2025 · similarity 0.85)
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.