Research map / Vision-language and multimodal / Human and embodied modeling
High-contrast “gaudy” images improve the training of deep neural network models of visual cortex
Read the paperCheck 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 |
|---|---|---|---|
| Train on a mix of gaudy and normal images instead of only normal images | Readout DNN predicting simulated visual-cortex responses · Simulated V4 responses, heldout normal images | fraction variance explained (R²) | Helped |
| Use preselected gaudy images instead of active-learning-selected images | Ensemble readout DNN · Simulated V4 responses, heldout normal images | fraction variance explained (R²) | Helped |
| Train on gaudy rather than normal images | GLM with linear, ReLU, or sigmoid activation · Gabor GLM responses, heldout normal images | fraction variance explained (R²) | Helped |
| Replace ReLU readout activations with sigmoid activations (ablation) | Readout DNN predicting simulated visual-cortex responses · Simulated V4 responses, heldout normal images | fraction variance explained (R²) | Helped |
| Increase readout depth to 10 layers (ablation) | Readout DNN predicting simulated visual-cortex responses · Simulated V4 responses, heldout normal images | fraction variance explained (R²) | Helped |
| Use a ResNet-block readout architecture (ablation) | Readout DNN predicting simulated visual-cortex responses · Simulated V4 responses, heldout normal images | fraction variance explained (R²) | Helped |
| Replace the nonlinear readout with a linear mapping (ablation) | Readout model predicting simulated visual-cortex responses · Simulated V4 responses, heldout normal images | fraction variance explained (R²) | No clear effect |
| Use only gaudy images instead of mixing gaudy and normal images (ablation) | Readout DNN predicting simulated visual-cortex responses · Simulated V4 responses, heldout normal images | fraction variance explained (R²) | Hurt |
| Smooth gaudy images to reduce edge contrast (ablation) | 3-layer ReLU readout network · Simulated V4 responses, heldout normal images | fraction variance explained (R²) | Hurt |
| Transform only high-edge-intensity pixels to gaudy (ablation) | 3-layer ReLU readout network · Simulated V4 responses, heldout normal images | fraction variance explained (R²) | Helped |
| Transform low-edge-intensity pixels rather than high-edge-intensity pixels (ablation) | 3-layer ReLU readout network · Simulated V4 responses, heldout normal images | fraction variance explained (R²) | Mixed |
| Smooth normal images before applying the gaudy transformation, removing texture while retaining edges (ablation) | 3-layer ReLU readout network · Simulated V4 responses, heldout normal images | fraction variance explained (R²) | No clear effect |
Closest papers by meaning
- Performance-optimized deep neural networks are evolving into worse models of inferotemporal visual cortex (NeurIPS 2023 · similarity 0.88)
- Benchmarking Out-of-Distribution Generalization Capabilities of DNN-based Encoding Models for the Ventral Visual Cortex. (NeurIPS 2024 · similarity 0.86)
- Generalization in data-driven models of primary visual cortex (ICLR 2021 · similarity 0.86)
- Towards robust vision by multi-task learning on monkey visual cortex (NeurIPS 2021 · similarity 0.86)
- Explicitly Modeling Subcortical Vision with a Neuro-Inspired Front-End Improves CNN Robustness (NeurIPS 2025 · similarity 0.84)
- L-WISE: Boosting Human Visual Category Learning Through Model-Based Image Selection and Enhancement (ICLR 2025 · similarity 0.84)
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.