Research map / Detection and segmentation / CNN features
Inlier-Centric Post-Training Quantization for Object Detection Models
Minsu Kim, Dongyeun Lee, Jaemyung Yu, Jiwan Hur, Giseop Kim, Junmo Kim
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 heatmap top-K selection to anomaly-only quantization (ablation) | InlierQ · COCO 2D camera W4A4 | mAP: 32.5 -> 34.5 | Improved |
| add heatmap top-K selection to anomaly-only quantization (ablation) | InlierQ · nuScenes 3D camera W4A4 | mAP: 24.8 -> 25.8 | Improved |
| add heatmap top-K selection to anomaly-only quantization (ablation) | InlierQ · nuScenes 3D LiDAR W4A4 | mAP: 44.2 -> 45.7 | Improved |
| restrict quantization to inliers instead of anomalies (with heatmap) (ablation) | InlierQ · COCO 2D camera W4A4 | mAP: 34.5 -> 34.7 | Improved |
| exclude anomalies from inlier set (ablation) | InlierQ · COCO 2D camera W4A4 | mAP: 34.6 -> 34.7 | Improved |
| exclude anomalies from inlier set (ablation) | InlierQ · nuScenes 3D camera W4A4 | mAP: 26.1 -> 26.4 | Improved |
| exclude anomalies from inlier set (ablation) | InlierQ · nuScenes 3D LiDAR W4A4 | mAP: 46.0 -> 46.6 | Improved |
| use EM instead of K-Means for inlier-anomaly clustering (ablation) | InlierQ · nuScenes 3D LiDAR W4A4 | mAP: 43.5 -> 46.6 | Improved |
| use EM instead of SVM for inlier-anomaly clustering (ablation) | InlierQ · nuScenes 3D LiDAR W4A4 | mAP: 46.6 -> 46.6 | No clear effect |
| raise heatmap top-K beyond training-time K (ablation) | InlierQ · nuScenes 3D detection | mAP | Hurt |
| use InlierQ instead of LiDAR-PTQ at W4A4 | LiDAR-PTQ · nuScenes 3D LiDAR CenterPoint | mAP: 39.5 -> 46.6 | Improved |
| use InlierQ instead of BRECQ at 8-bit weights and activations | BRECQ · COCO RetinaNet W8A8 | mAP: 38.9 -> 39.0 | No clear effect |
Closest papers by meaning
- A Likelihood Ratio-Based Approach to Segmenting Unknown Objects (IJCV 2025 · similarity 0.87)
- Normalizing Flow Based Feature Synthesis for Outlier-Aware Object Detection (CVPR 2023 · similarity 0.86)
- Quantile-Based Maximum Likelihood Training for Outlier Detection (AAAI 2024 · similarity 0.86)
- Reg-PTQ: Regression-specialized Post-training Quantization for Fully Quantized Object Detector (CVPR 2024 · similarity 0.86)
- Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models (NeurIPS 2025 · similarity 0.85)
- LiDAR-PTQ: Post-Training Quantization for Point Cloud 3D Object Detection (ICLR 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.
