Research map / Detection and segmentation / CNN features
Delving Into Salient Object Subitizing and Detection
Shengfeng He, Jianbo Jiao, Xiaodan Zhang, Guoqiang Han, Rynson W.H. Lau
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 subitizing-driven adaptive weight layer versus independent detection network | salient object detection network · Pascal-S | F-measure: 0.721 -> 0.794 | Improved |
| add subitizing-driven adaptive weight layer versus independent detection network | salient object detection network · Pascal-S | MAE: 0.143 -> 0.114 | Improved |
| adaptive weight layer versus shared multi-task network | multi-task saliency network · Pascal-S | F-measure: 0.770 -> 0.794 | Improved |
| adaptive weight layer versus concatenated network | multi-task saliency network · Pascal-S | MAE: 0.125 -> 0.114 | Improved |
| adaptive weight layer versus cascaded network | multi-task saliency network · Pascal-S | F-measure: 0.753 -> 0.794 | Improved |
| share Conv5 only between subitizing and detection (ablation) | shared network · MSO | detection MAE (difference vs independent) and subitizing mAP: -0.53 MAE; +4.17 mAP | Improved |
| share Conv3-5 between subitizing and detection (ablation) | shared network · MSO | detection MAE (difference) and subitizing mAP: +0.42 MAE; -3.56 mAP | Mixed |
| share Conv1-5 (all layers) between subitizing and detection (ablation) | shared network · MSO | detection MAE (difference) and subitizing mAP: -0.43 MAE; -0.62 mAP | Mixed |
| vary number of shared layers (ablation) | shared network · MSO | detection and subitizing | Mixed |
| add subitizing guidance to unconstrained detection versus MAP | adaptive weight network · MSO (full) | adaptive precision: 0.370 -> 0.654 | Improved |
| add subitizing guidance on background-only images versus MAP | adaptive weight network · MSO (background images only) | MAE: 0.068 -> 0.061 | Improved |
| swap tasks so that detection guides subitizing, versus MSO baseline | adaptive weight network · Subitizing, 4+ objects | average precision (%): 79.0% -> 84.1% | 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.
- RFCN — R-FCN: Object Detection via Region-based Fully Convolutional Networks (NeurIPS 2016) · better on 5 of 6 shared results
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
- Salient Object Subitizing (IJCV 2017 · similarity 0.91)
- Revisiting Salient Object Detection: Simultaneous Detection, Ranking, and Subitizing of Multiple Salient Objects (CVPR 2018 · similarity 0.91)
- Recurrently Aggregating Deep Features for Salient Object Detection (AAAI 2018 · similarity 0.89)
- Reverse Attention for Salient Object Detection (ECCV 2018 · similarity 0.89)
- Salient Object Detection With Pyramid Attention and Salient Edges (CVPR 2019 · similarity 0.89)
- Multi-Scale Interactive Network for Salient Object Detection (CVPR 2020 · similarity 0.89)
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.
