Research map / Detection and segmentation / 3D detection and driving
Feature Compression for Cloud-Edge Multimodal 3D Object Detection
Chongzhen Tian, Zhengxin Li, Hui Yuan, Raouf Hamzaoui, Liquan Shen, Sam Kwong
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 masked transformer enhancement (ablation) | T-FFC · KITTI val Car 3D AP Hard | AP: 77.8811 -> 82.8727 | Improved |
| add masked transformer enhancement (ablation) | T-FFC · KITTI val Car BEV AP Hard | AP: 86.9797 -> 89.4281 | Improved |
| add decoder convolution enhancement (ablation) | T-FFC · KITTI val Car BEV AP Easy | AP: 93.7519 -> 96.3852 | Improved |
| add decoder convolution enhancement (ablation) | T-FFC · KITTI val Car 3D AP Easy | AP: 93.0963 -> 92.9419 | Hurt |
| add decoder convolution enhancement (ablation) | A-FFC · KITTI val Car 3D AP Hard | AP: 81.1472 -> 85.4900 | Improved |
| add channel compression of newly added features F3 and F4 (ablation) | A-FFC · KITTI val Car 3D AP Hard | AP: 85.5285 -> 85.4900 | No clear effect |
| compress features with T-FFC versus uncompressed network | VirConv-L · KITTI val Car 3D AP Hard | AP: 85.4861 -> 82.8727 | Hurt |
| compress features with A-FFC versus uncompressed network | VirConv-L · KITTI val Car 3D AP Hard | AP: 85.4861 -> 85.4900 | No clear effect |
| use A-FFC mode instead of T-FFC mode (extra F3 and F4 features) | FFC · KITTI val Car 3D AP Hard | AP: 82.8727 -> 85.4900 | Improved |
| increase post-processing distance d_p from 2 to 6 (ablation) | 3D object reconstruction · KITTI val Chamfer Distance | Chamfer Distance: 0.012967 -> 0.009470 | Improved |
| reduce surface judgment threshold d_s from 0.4 to 0.2 (U_c^d 1 to 2) (ablation) | 3D object reconstruction · KITTI val Chamfer Distance | Chamfer Distance: 0.011568 -> 0.007909 | Improved |
Closest papers by meaning
- Temporal and Spatial Representation Learning for Multimodal Low-Beam 3D Object Detection (AAAI 2026 · similarity 0.86)
- Fully Sparse 3D Object Detection (NeurIPS 2022 · similarity 0.86)
- Revisiting Token Compression for Accelerating ViT-based Sparse Multi-View 3D Object Detectors (CVPR 2026 · similarity 0.86)
- ObjectFusion: Multi-modal 3D Object Detection with Object-Centric Fusion (ICCV 2023 · similarity 0.85)
- SAFDNet: A Simple and Effective Network for Fully Sparse 3D Object Detection (CVPR 2024 · similarity 0.84)
- FSHNet: Fully Sparse Hybrid Network for 3D Object Detection (CVPR 2025 · similarity 0.84)
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.
