Research map / Detection and segmentation / CNN features
BlitzNet: A Real-Time Deep Network for Scene Understanding
Nikita Dvornik, Konstantin Shmelkov, Julien Mairal, Cordelia Schmid
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 joint segmentation training to detection (ablation) | BlitzNet300 · VOC12 val-seg | detection mAP (%): 78.9 -> 80.0 | Improved |
| add joint detection training to segmentation (ablation) | BlitzNet300 · VOC12 val-seg | segmentation mIoU (%): 72.4 -> 72.8 | Improved |
| add extra detection-only images (VOC07 trainval) to joint training (ablation) | BlitzNet300 · VOC12 val-seg | segmentation mIoU (%): 72.4 -> 75.7 | Improved |
| add extra detection-only images (VOC07 trainval) to joint training (ablation) | BlitzNet300 · VOC12 val-seg | detection mAP (%): 83.0 -> 83.6 | Improved |
| add joint segmentation training (ablation) | BlitzNet512 · COCO minival2014 | detection mAP (%): 33.2 -> 34.1 | Improved |
| add joint detection training (ablation) | BlitzNet512 · COCO minival2014 | segmentation mIoU (%): 48.3 -> 53.5 | Improved |
| add segmentation training to stride-4 model (ablation) | BlitzNet300 (s4) · VOC07 test | mAP (%): 78.2 -> 79.1 | Improved |
| add segmentation training to stride-8 model (ablation) | BlitzNet512 (s8) · VOC07 test | mAP (%): 80.7 -> 81.5 | Improved |
| use upscale-stream stride 4 instead of 8 (ablation) | BlitzNet300 · VOC07 test | mAP (%): 78.5 -> 78.2 | No clear effect |
| pretrain on COCO | BlitzNet512 · VOC12 test | mAP (%): 79.0 -> 83.8 | Improved |
| use ResSkip instead of hourglass-style fusion block (ablation) | BlitzNet · VOC07 test / VOC12 val | mAP and mIoU (%): 78.7 -> 79.1 | Improved |
| use ResSkip instead of refine-style fusion block (ablation) | BlitzNet · VOC07 test / VOC12 val | mAP and mIoU (%): 78.0 -> 79.1 | Mixed |
Later papers that compared with this method
1 later accepted paper reports BlitzNet300 as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- Triply Supervised Decoder Networks for Joint Detection and Segmentation (CVPR 2019 · better on 4 of 4 shared results)
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- R-FCN — R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016) · better on 2 of 2 shared results
- Faster R-CNN — Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks (TPAMI 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
- In Defense of Pre-Trained ImageNet Architectures for Real-Time Semantic Segmentation of Road-Driving Images (CVPR 2019 · similarity 0.89)
- Triply Supervised Decoder Networks for Joint Detection and Segmentation (CVPR 2019 · similarity 0.87)
- Full-Resolution Residual Networks for Semantic Segmentation in Street Scenes (CVPR 2017 · similarity 0.87)
- Lightweight and Progressively-Scalable Networks for Semantic Segmentation (IJCV 2023 · similarity 0.87)
- DeepVoting: A Robust and Explainable Deep Network for Semantic Part Detection Under Partial Occlusion (CVPR 2018 · similarity 0.86)
- Seamless Scene Segmentation (CVPR 2019 · 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.
