Research map / Detection and segmentation / CNN features
Highly Efficient Salient Object Detection with 100K Parameters
Shang-Hua Gao, Yong-Qiang Tan, Ming-Ming Cheng, Chengze Lu, Yunpeng Chen, Shuicheng Yan
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 |
|---|---|---|---|
| use 3/1 high/low channel split instead of 1/0 in extractor (ablation) | Extractor · ECSSD | max F-measure: 88.2 -> 88.6 | Improved |
| shift all channels to low resolution (0/1) instead of 1/0 in extractor (ablation) | Extractor · ECSSD | max F-measure: 88.2 -> 86.4 | Mixed |
| add cross-stages fusion gOctConvs to the extractor (ablation) | CSNet 5/5 · ECSSD | max F-measure: 88.6 -> 90.0 | Improved |
| use extreme low-resolution-only split 0/1 with cross-stages fusion (ablation) | CSNet · ECSSD | max F-measure: 90.0 -> 88.2 | Mixed |
| replace fixed 5/5 split with learnable channels (x2-L) | CSNet · ECSSD | max F-measure: 90.0 -> 91.6 | Improved |
| replace fixed 5/5 split with learnable channels (x1-L) | CSNet · ECSSD | max F-measure: 90.0 -> 90.0 | No clear effect |
| add dynamic weight decay to Pruning Filters (ablation) | Pruning Filters pruning · ECSSD | max F-measure: 88.7 -> 89.4 | Improved |
| add dynamic weight decay to Geometric-Median pruning (ablation) | Geometric-Median pruning · ECSSD | max F-measure: 88.7 -> 89.6 | Improved |
| add dynamic weight decay to BatchNorm weights during training (ablation) | CSNet-L · ECSSD | test MAE / convergence | Improved |
| increase initial channel width from x1 to x2.0 before pruning (ablation) | CSNet-L · ECSSD | max F-measure: 90.0 -> 91.6 | Improved |
| prune the width x2.0 model (ablation) | CSNet · ECSSD | max F-measure: 91.5 -> 91.6 | No clear effect |
| add CSF part to ResNet50 versus PoolNet decoder | ResNet50 · ECSSD | max F-measure: .940 -> .940 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 4 of 4 on most shared results.
- ESPNetv2 — ESPNet: Efficient Spatial Pyramid of Dilated Convolutions for Semantic Segmentation (ECCV 2018) · better on 2 of 2 shared results
- PiCANet — PiCANet: Learning Pixel-Wise Contextual Attention for Saliency Detection (CVPR 2018) · better on 1 of 1 shared result
- BiseNet — BiSeNet: Bilateral Segmentation Network for Real-time Semantic Segmentation (ECCV 2018) · better on 1 of 1 shared result
- Amulet — Amulet: Aggregating Multi-Level Convolutional Features for Salient Object Detection (ICCV 2017) · 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
- SALISA: Saliency-Based Input Sampling for Efficient Video Object Detection (ECCV 2022 · similarity 0.87)
- Weakly Supervised Salient Object Detection Using Image Labels (AAAI 2018 · similarity 0.87)
- Breaking Barriers, Localizing Saliency: A Large-Scale Benchmark and Baseline for Condition-Constrained Salient Object Detection (TPAMI 2025 · similarity 0.87)
- Deep Embedding Features for Salient Object Detection (AAAI 2019 · similarity 0.87)
- Suppress and Balance: A Simple Gated Network for Salient Object Detection (ECCV 2020 · similarity 0.86)
- Cascaded Partial Decoder for Fast and Accurate Salient Object Detection (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.
