Layer-Wise Relevance Propagation Heatmaps for a human-aligned CNN

This dataset contains Layer-wise Relevance Propagation (LRP) heatmaps for the VGG-Face Human Judgment model from Hofmann et al. (2024). The model is a human-aligned VGG-Face (Parkhi et al., 2015) architecture fine-tuned by Hofmann et al. (2024) on behavioral data to predict human odd-one-out choices in face triplet similarity judgments. The provided heatmaps highlight facial regions (e.g., eyes, nose, mouth) that contributed to the model’s predictions and reflect decision-relevant information aligned with human similarity judgments. The dataset includes: Relevance maps as .npy files for each odd-one-out prediction of a face triplet. Visual previews of each relevance map in .png format. Metadata CSV mapping each heatmap to triplet indices, human ground truth, and model predictions. Heatmap ComputationHeatmaps were computed using Layer-wise Relevance Propagation (Bach et al., 2015) implemented with the Zennit library (Anders et al., 2021), using an Epsilon–Alpha2Beta1–Flat composite. Relevance was propagated from the model’s predicted odd-one-out class, and relevance values were summed across RGB channels to obtain 2D maps. Visualization PNG previews use a symmetric seismic colormap, normalized to the maximum absolute relevance value of each individual map ($v_{min} = -|max|$, $v_{max} = |max|$), where red indicates positive relevance (supporting the decision) and blue indicates negative relevance. Please read the README_HUM.txt for more information about how to use the data. Source code for model training and heatmap generation is available via the linked GitHub repository FaceSimXAI.