Visualise the patch-level feature space of ViT-S/16 models from the OK-AI ImageNet-1K SSL collection.
Upload any image and the app runs it through all three models, applies PCA to the patch token representations, and renders the first 3 principal components as an RGB image.
Visualise the patch-level feature space of ViT-S/16 models from the OK-AI ImageNet-1K SSL collection.
Upload any image and the app runs it through all three models, applies PCA to the patch token representations, and renders the first 3 principal components as an RGB image.
