Research map / Detection and segmentation / CNN features
Single-Shot Object Detection With Enriched Semantics
Zhishuai Zhang, Siyuan Qiao, Cihang Xie, Wei Shen, Bo Wang, Alan L. Yuille
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 DES semantic enrichment to SSD300 | SSD300 (VGG16) · VOC2007 test | mAP: 77.5 to 79.7 | Improved |
| Add DES semantic enrichment to SSD512 | SSD512 (VGG16) · VOC2007 test | mAP: 79.5 to 81.7 | Improved |
| Add DES semantic enrichment to SSD300 | SSD300 (VGG16) · COCO test-dev | mAP: 25.1 to 28.3 | Improved |
| Add DES semantic enrichment to SSD512 | SSD512 (VGG16) · COCO test-dev | mAP: 28.8 to 32.8 | Improved |
| Add the global activation module (ablation) | SSD300 · VOC2007 test | mAP: 77.5 -> 78.1 | Improved |
| Add the segmentation branch with α=0.1 to the global-activation model (ablation) | SSD300+G · VOC2007 test | mAP: 78.1 -> 79.7 | Improved |
| Use segmentation loss weight α=0.1 instead of α=0.0 (ablation) | SSD300+G+S · VOC2007 test | mAP: 79.4 -> 79.7 | Improved |
| Increase segmentation loss weight from α=0.1 to α=1.0 (ablation) | SSD300+G+S · VOC2007 test | mAP: 79.7 -> 78.6 | Hurt |
| Train segmentation and detection in parallel instead of activating detection features with segmentation features (ablation) | SSD300+G+S · VOC2007 test | mAP: 79.7 -> 78.2 | Hurt |
| Replace segmentation supervision and feature activation with a deeper VGG16-style detector (ablation) | SSD300+G+S · VOC2007 test | mAP: 79.7 -> 77.6 | Hurt |
| Use DES instead of SSD512 | SSD512 (VGG16) · VOC2007 test | time (ms/img): 18.6 -> 31.5 | Hurt |
Closest papers by meaning
- UniT: Unified Knowledge Transfer for Any-Shot Object Detection and Segmentation (CVPR 2021 · similarity 0.88)
- Enriched Feature Guided Refinement Network for Object Detection (ICCV 2019 · similarity 0.88)
- Learning Rich Features at High-Speed for Single-Shot Object Detection (ICCV 2019 · similarity 0.88)
- Semantic Relation Reasoning for Shot-Stable Few-Shot Object Detection (CVPR 2021 · similarity 0.87)
- DesCo: Learning Object Recognition with Rich Language Descriptions (NeurIPS 2023 · similarity 0.87)
- Unlocking the Potential of Pre-trained Vision Transformers for Few-Shot Semantic Segmentation through Relationship Descriptors (CVPR 2024 · similarity 0.87)
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.
