Research map / Detection and segmentation / CNN features
Ubernet: Training a Universal Convolutional Neural Network for Low-, Mid-, and High-Level Vision Using Diverse Datasets and Limited Memory
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 object detection as a second task to segmentation (ablation) | UberNet · PASCAL VOC 2012 test | mean IoU: 72.4 -> 72.3 | No clear effect |
| Train segmentation jointly with seven tasks instead of one (ablation) | UberNet · PASCAL VOC 2012 test | mean IoU: 72.4 -> 68.7 | Hurt |
| Train human-part segmentation jointly with seven tasks instead of one (ablation) | UberNet · Human part segmentation dataset | mean IoU: 51.98 -> 48.82 | Hurt |
| Train semantic boundary detection jointly with seven tasks instead of one (ablation) | UberNet · PASCAL VOC 2010 val | mAP: 54.3 -> 44.3 | Hurt |
| Train boundary detection jointly with seven tasks instead of one (ablation) | UberNet · BSD test | ODS: 0.791 -> 0.785 | Hurt |
| Train saliency estimation jointly with seven tasks instead of one (ablation) | UberNet · PASCAL-S | Maximal F-measure: 0.835 -> 0.823 | Hurt |
| Train object detection jointly with seven tasks instead of one (ablation) | UberNet · PASCAL VOC 2007 test | mAP: 78.7 -> 77.8 | Hurt |
| Add semantic segmentation as a second task to detection (ablation) | UberNet · PASCAL VOC 2007 test | mAP: 78.7 -> 80.1 | Helped |
| Raise the normal-estimation loss weight from γ=1 to γ=50 (ablation) | UberNet, 1-Task · NYU-v2 | Mean angle: 23.9 -> 21.4 | Helped |
| Raise the normal-estimation loss weight from γ=1 to γ=10 (ablation) | UberNet, 1-Task · NYU-v2 | Mean angle: 23.9 -> 23.2 | Helped |
| Raise the normal-estimation loss weight from γ=1 to γ=5 (ablation) | UberNet, 1-Task · NYU-v2 | Mean angle: 23.9 -> 23.3 | Helped |
| Increase normal-estimation loss weight from γ=1 to γ=50 (ablation) | UberNet, 7-Task · Seven-task results | Detection mAP: 77.8 -> 73.5 | Hurt |
Closest papers by meaning
- Rethinking the Inception Architecture for Computer Vision (CVPR 2016 · similarity 0.89)
- SCNN: A General Distribution Based Statistical Convolutional Neural Network with Application to Video Object Detection (AAAI 2019 · similarity 0.88)
- Fast Neural Architecture Search of Compact Semantic Segmentation Models via Auxiliary Cells (CVPR 2019 · similarity 0.88)
- RePr: Improved Training of Convolutional Filters (CVPR 2019 · similarity 0.88)
- Deep Layer Aggregation (CVPR 2018 · similarity 0.88)
- AsCAN: Asymmetric Convolution-Attention Networks for Efficient Recognition and Generation (NeurIPS 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.