Research map / Detection and segmentation / CNN features
Object Guided External Memory Network for Video Object Detection
Hanming Deng, Yang Hua, Tao Song, Zongpu Zhang, Zhengui Xue, Ruhui Ma, Neil Robertson, Haibing Guan
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 pixel-level object-guided memory Mpix (ablation) | R-FCN · ImageNet VID val | mAP (%): 73.8 -> 78.8 | Improved |
| add instance-level memory Minst (ablation) | R-FCN · ImageNet VID val | mAP (%): 73.8 -> 74.8 | Improved |
| add instance memory Minst to pixel memory (ablation) | Mpix · ImageNet VID val | mAP (%): 78.8 -> 79.3 | Improved |
| add pixel memory Mpix (ablation) | R-FCN · ImageNet VID val (fast motion) | mAP (%): 52.4 -> 60.0 | Improved |
| add instance memory to pixel memory (ablation) | Mpix · ImageNet VID val (slow motion) | mAP (%): 85.8 -> 86.2 | Improved |
| add both memories (ablation) | R-FCN · ImageNet VID val | runtime (ms): 72 -> 112 | Hurt |
| down-sample non-key frames to 1/2 with memory (ablation) | Ours · ImageNet VID val | mAP (%): 73.8 -> 76.8 | Improved |
| down-sample non-key frames to 1/3 with memory (ablation) | Ours · ImageNet VID val | mAP (%): 73.8 -> 74.4 | Improved |
| down-sample non-key frames to 1/4 with memory (ablation) | Ours · ImageNet VID val | mAP (%): 73.8 -> 71.5 | Hurt |
| add deformable convolution (DCN) to feature network | Ours · ImageNet VID val | mAP (%): 79.3 -> 80.0 | Improved |
| apply Seq-NMS post-processing | Ours · ImageNet VID val | mAP (%): 79.3 -> 80.8 | Improved |
| replace flow-based alignment with attentional read (vs FGFA) | Ours · ImageNet VID val (fast motion) | mAP (%): 57.6 -> 61.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- R-FCN — R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Video Object Detection with an Aligned Spatial-Temporal Memory (ECCV 2018 · similarity 0.88)
- Mobile Video Object Detection With Temporally-Aware Feature Maps (CVPR 2018 · similarity 0.88)
- MAMBA: Multi-level Aggregation via Memory Bank for Video Object Detection (AAAI 2021 · similarity 0.87)
- Memory Enhanced Global-Local Aggregation for Video Object Detection (CVPR 2020 · similarity 0.87)
- Fast Object Detection in Compressed Video (ICCV 2019 · similarity 0.87)
- Context Enhanced Transformer for Single Image Object Detection in Video Data (AAAI 2024 · similarity 0.86)
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.
