Research map / Detection and segmentation / CNN features
Towards Persistence: Learning Topological Constraints for Event-based Small Object Detection
Shiman He, Nuo Chen, Xinyi Ying, Yihang Luo, Yangsi Shi, Zaiping Lin, Miao Li
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 |
|---|---|---|---|
| add TLM to the baseline (ablation) | SpTopoNet baseline · EV-UAV | IoU (%): 52.77 -> 63.32 | Improved |
| add SCM to the baseline (ablation) | SpTopoNet baseline · EV-UAV | IoU (%): 52.77 -> 62.50 | Improved |
| add EvTopoLoss to the baseline alone (ablation) | SpTopoNet baseline · EV-UAV | IoU (%): 52.77 -> 61.47 | Improved |
| add EvTopoLoss on top of TLM + SCM (ablation) | SpTopoNet · EV-UAV | IoU (%): 64.07 -> 66.62 | Improved |
| place TLM in all three encoder layers instead of none (ablation) | SpTopoNet · EV-UAV | IoU (%): 60.64 -> 66.62 | Improved |
| place TLM only in layer 1 vs layers 2 or 3 (ablation) | SpTopoNet · EV-UAV | IoU (%): layer 1 64.80 vs layer 2 61.57, layer 3 61.60 | Improved |
| add EvTopoLoss to Ev-SpSegNet (ablation) | Ev-SpSegNet · EV-UAV | IoU (%): 53.62 -> 62.11 | Improved |
| add EvTopoLoss to 3D-UNet (ablation) | 3D-UNet · EV-UAV | IoU (%): 48.32 -> 53.22 | Improved |
| add EvTopoLoss to PointNet++ (ablation) | PointNet++ · EV-UAV | IoU (%): 35.44 -> 36.32 | Improved |
| set topology loss weight lambda_topo to 0.05 instead of 0.01 (ablation) | SpTopoNet · EV-UAV validation | IoU (%): 58.33 -> 67.17 | Improved |
| raise lambda_topo from 0.05 to 0.1 (ablation) | SpTopoNet · EV-UAV validation | IoU (%): 67.17 -> 54.19 | Hurt |
| SpTopoNet vs prior best point-cloud method Ev-SpSegNet | SpTopoNet · EV-UAV test | runtime (ms): 35.9 -> 56.5 | Hurt |
Closest papers by meaning
- Event-based Tiny Object Detection: A Benchmark Dataset and Baseline (ICCV 2025 · similarity 0.91)
- Following the Flow: Advection-Consistent Modeling for Event-based Small Object Detection (ECCV 2026 · similarity 0.88)
- Low-latency Event-based Object Detection with Spatially-Sparse Linear Attention (ECCV 2026 · similarity 0.87)
- SMamba: Sparse Mamba for Event-based Object Detection (AAAI 2025 · similarity 0.86)
- PASS: Path-selective State Space Model for Event-based Recognition (NeurIPS 2025 · similarity 0.86)
- Scene Adaptive Sparse Transformer for Event-based Object Detection (CVPR 2024 · 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.
