Research map / Probabilistic and scientific modeling / Variational and generative sampling
An Energy-Based Prior for Generative Saliency
Jing Zhang, Jianwen Xie, Nick M. Barnes, Ping Li
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 GAN latent variable to deterministic base (ablation) | Swin saliency model · DUTS test | S-measure: .901 -> .911 | Improved |
| add VAE latent variable to deterministic base (ablation) | Swin saliency model · DUTS test | S-measure: .901 -> .908 | Improved |
| add ABP Gaussian latent variable to deterministic base (ablation) | ResNet50 saliency model · DUTS test | S-measure: .884 -> .876 | Hurt |
| add ABP Gaussian latent variable to deterministic base (ablation) | VGG16 saliency model · DUTS test | S-measure: .867 -> .859 | Hurt |
| replace Gaussian prior with EBM prior in ABP (EABP vs ABP) (ablation) | Swin saliency model · DUTS test | MAE: .029 -> .029 | No clear effect |
| replace Gaussian prior with EBM prior in ABP (EABP vs ABP) (ablation) | ResNet50 saliency model · DUTS test | S-measure: .876 -> .877 | Improved |
| add multi-scale training augmentation to EABP (ablation) | ResNet50 EABP · DUTS test | S-measure: .877 -> .896 | Improved |
| add multi-scale training augmentation to EABP (ablation) | VGG16 EABP · DUTS test | S-measure: .870 -> .876 | Improved |
| use Swin backbone instead of ResNet50 (ablation) | EABP · DUTS test | S-measure: .877 -> .908 | Improved |
| use ResNet50 instead of VGG16 backbone (ablation) | EABP · DUTS test | S-measure: .870 -> .877 | Improved |
| apply EBM-prior generative framework to deterministic VST | VST · DUTS test | S-measure: .896 -> .895 | No clear effect |
| apply EGAN to VST for RGB-D | VST · LFSD (RGB-D) | S-measure: .890 -> .882 | Hurt |
Closest papers by meaning
- Learning Generative Vision Transformer with Energy-Based Latent Space for Saliency Prediction (NeurIPS 2021 · similarity 0.96)
- Energy-Based Generative Cooperative Saliency Prediction (AAAI 2022 · similarity 0.93)
- Uncertainty Inspired RGB-D Saliency Detection (TPAMI 2021 · similarity 0.90)
- Learning Multimodal Latent Generative Models with Energy-Based Prior (ECCV 2024 · similarity 0.87)
- Saliency Hierarchy Modeling via Generative Kernels for Salient Object Detection (ECCV 2022 · similarity 0.87)
- Learning normalized image densities via dual score matching (NeurIPS 2025 · similarity 0.85)
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.
