concepts.vision.fm_match.dino.get_feature

Functions

compute_dino_feature(source_img[, ...])	return: (result, resized_imgs, downsampled_imgs), where result is a tensor of shape (N, pca_dim, num_patches, num_patches),
resize(img, target_res[, resize, to_pil, edge])
torch_pca(feature[, target_dim])	Perform Principal Component Analysis (PCA) on the input feature tensor.