FeatLens: Model-agnostic feature-map visualization
PCA-to-RGB feature maps from any vision model and layer — DINO, DINOv2/v3, CLIP, SigLIP, MAE, DeiT, V-JEPA, CNNs, and more via timm or Hugging Face.
Model-agnostic feature-map visualization for visual foundation models. FeatLens renders dense feature maps from any vision backbone and any layer as a model × layer grid, with PCA, cosine-similarity seeds, k-means, foreground masks, saliency, cross-image correspondence, and attention rollout.