Research map / Point clouds / Segmentation and detection
ESAM++: Efficient Online 3D Perception on the Edge
Qin Liu, Lavisha Aggarwal, Saptarashmi Bandyopadhyay, Vikas Bahirwani, Marc Niethammer, Ehsan Adeli, Andrea Colaco
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 |
|---|---|---|---|
| use SFPN-Small instead of MinkNet18 backbone | ESAM++ · ScanNet200 CPU | mAP: 27.1 -> 30.3 | Improved |
| use SFPN-Small instead of MinkNet18 backbone | ESAM++ · ScanNet200 CPU | CPU latency (ms): 487ms -> 211ms | Improved |
| use SFPN-Small instead of SPVNAS | ESAM++ · ScanNet200 CPU | mAP: 22.8 -> 30.3 | Improved |
| use SFPN-Large instead of ESAM-E Sparse UNet | ESAM++ · ScanNet200 CPU | CPU latency (ms): 934ms -> 326ms | Improved |
| use SFPN-Large instead of ESAM-E Sparse UNet | ESAM++ · ScanNet200 | mAP: 43.4 -> 43.7 | No clear effect |
| scale backbone from SFPN-Small to SFPN-Base (ablation) | ESAM++ · ScanNet200 | mAP: 30.3 -> 39.7 | Improved |
| scale backbone from SFPN-Base to SFPN-Large (ablation) | ESAM++ · ScanNet200 | mAP: 39.7 -> 43.7 | Improved |
| train with 100% instead of 50% of the data (ablation) | ESAM++ · ScanNet200 class-agnostic | AP: 40.6 -> 43.7 | Improved |
| train with 50% instead of 10% of the data (ablation) | ESAM++ · ScanNet200 class-agnostic | AP: 32.4 -> 40.6 | Improved |
Closest papers by meaning
- EmbodiedSAM: Online Segment Any 3D Thing in Real Time (ICLR 2025 · similarity 0.89)
- ODIN: A Single Model for 2D and 3D Segmentation (CVPR 2024 · similarity 0.87)
- Memory-based Adapters for Online 3D Scene Perception (CVPR 2024 · similarity 0.87)
- Online Segment Any 3D Thing as Instance Tracking (NeurIPS 2025 · similarity 0.87)
- EA3D: Online Open-World 3D Object Extraction from Streaming Videos (NeurIPS 2025 · similarity 0.86)
- Edge-Aware 3D Instance Segmentation Network with Intelligent Semantic Prior (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.
