Research map / Architectures and efficiency / CNN design
FovEx: Human-Inspired Explanations for Vision Transformers and Convolutional Neural Networks
Mahadev Prasad Panda, Matteo Tiezzi, Martina G. Vilas, Gemma Roig, Bjoern M. Eskofier, Dario Zanca
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 random restart (rr=True) to fixation search (ablation) | FovEx ViT-B/16 · ImageNet-1K val | Avg. % drop: 17.561 -> 13.973 | Improved |
| add random restart (rr=True) to fixation search (ablation) | FovEx ViT-B/16 · ImageNet-1K val | EBPG: 51.067 -> 47.705 | Hurt |
| add random restart (rr=True) to fixation search (ablation) | FovEx ViT-B/16 · ImageNet-1K val | Insert: 0.824 -> 0.840 | Improved |
| add random restart (rr=True) to fixation search (ablation) | FovEx ViT-B/16 · ImageNet-1K val | Avg. % increase: 26.389 -> 30.389 | Improved |
| increase foveation sigma sigma_f (ablation) | FovEx ViT-B/16 · ImageNet-1K val | Avg. % drop / Avg. % increase / Delete | Mixed |
| increase scanpath length N (ablation) | FovEx ViT-B/16 · ImageNet-1K val | quality metrics and time | Mixed |
| increase optimization steps os (ablation) | FovEx ViT-B/16 · ImageNet-1K val | quality metrics and time | Mixed |
| increase blur sigma sigma_b (ablation) | FovEx ViT-B/16 · ImageNet-1K val | quality metrics | Mixed |
| increase blur filter size bfs (ablation) | FovEx ViT-B/16 · ImageNet-1K val | quality metrics | Mixed |
| increase forgetting factor beta (ablation) | FovEx ViT-B/16 · ImageNet-1K val | explanation quality | Mixed |
| use FovEx instead of RISE | ViT-B/16 explainer · ImageNet-1K val | Avg. % drop: 15.673 -> 13.970 | Improved |
| use FovEx instead of gradCAM | ViT-B/16 explainer · MIT1003 human gaze | NSS: 0.2357 -> 0.7160 | Improved |
Closest papers by meaning
- Seeing More with Less: Human-like Representations in Vision Models (CVPR 2025 · similarity 0.88)
- Mind the GAP: Glimpse-based Active Perception improves generalization and sample efficiency of visual reasoning (ICLR 2025 · similarity 0.87)
- LLMind: Bio-inspired Training-free Adaptive Visual Representations for Vision-Language Models (CVPR 2026 · similarity 0.87)
- The All-Seeing Project: Towards Panoptic Visual Recognition and Understanding of the Open World (ICLR 2024 · similarity 0.87)
- On the Faithfulness of Vision Transformer Explanations (CVPR 2024 · similarity 0.87)
- Probabilistic Conceptual Explainers: Trustworthy Conceptual Explanations for Vision Foundation Models (ICML 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.
