Research map / Detection and segmentation / Tracking
Tracking the Untrackable: Learning to Track Multiple Cues With Long-Term Dependencies
Amir Sadeghian, Alexandre Alahi, Silvio Savarese
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 |
|---|---|---|---|
| replace target RNN with a fully-connected layer (ablation) | AMIR · MOT15 validation | MOTA: 18.2 -> 30.8 | Improved |
| use RNNs for all cue networks instead of fully-connected layers (ablation) | AMIR · MOT15 validation | MOTA: 12.9 -> 30.8 | Improved |
| add interaction cue to appearance and motion (ablation) | AMIR · MOT15 validation | MOTA: 28.8 -> 30.8 | Improved |
| add motion cue to appearance and interaction (ablation) | AMIR · MOT15 validation | MOTA: 27.4 -> 30.8 | Improved |
| add appearance cue to motion and interaction (ablation) | AMIR · MOT15 validation | MOTA: 22 -> 30.8 | Improved |
| combine three cues instead of appearance only (ablation) | AMIR · MOT15 validation | MOTA: 23.7 -> 30.8 | Improved |
| combine three cues instead of interaction only (ablation) | AMIR · MOT15 validation | MOTA: 15.4 -> 30.8 | Improved |
| increase LSTM sequence length (ablation) | AMIR · MOT15 validation | MOTA | Improved |
| increase LSTM sequence length on Stanford Drone (ablation) | AMIR · Stanford Drone validation | MOTA | No clear effect |
| replace MDP hand-crafted representation with learned multi-cue representation | MDP · MOT15 test | MOTA: 30.3 -> 37.6 | Improved |
| fine-tune on drone data (ablation) | AMIR · Stanford Drone Dataset | MOTA: 78.6 -> 82.9 | Improved |
| learned representation vs MDP + SF-mc on drone data | AMIR · Stanford Drone Dataset | MOTA: 75.6 -> 78.6 | Improved |
Closest papers by meaning
- Online Multi-Target Tracking Using Recurrent Neural Networks (AAAI 2017 · similarity 0.90)
- MUTrack: A Memory-Aware Unified Representation Framework for Visual Tracking (AAAI 2026 · similarity 0.89)
- Multi-object Tracking with Neural Gating Using Bilinear LSTM (ECCV 2018 · similarity 0.88)
- Recurrently Target-Attending Tracking (CVPR 2016 · similarity 0.88)
- Tracking objects that change in appearance with phase synchrony (ICLR 2025 · similarity 0.87)
- Track-On: Transformer-based Online Point Tracking with Memory (ICLR 2025 · 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.
