The basic idea of SPADE is relatively well understood, but some parts of it are not understood in the calculation of AnomalyMap:
for d_idx in range(feat_gallery.shape[0] // 100): # TODO:
feat_gallery_map = feat_gallery[d_idx 100:d_idx 100 + 100].to(self.device)
dist_matrix = torch.pairwise_distance(feat_gallery_map, test_feat_map.to(self.device))
dist_matrix_list.append(dist_matrix)
What is the meaning of 100 in this place? never understood it.
The basic idea of SPADE is relatively well understood, but some parts of it are not understood in the calculation of AnomalyMap: for d_idx in range(feat_gallery.shape[0] // 100): # TODO: feat_gallery_map = feat_gallery[d_idx 100:d_idx 100 + 100].to(self.device) dist_matrix = torch.pairwise_distance(feat_gallery_map, test_feat_map.to(self.device)) dist_matrix_list.append(dist_matrix)
What is the meaning of 100 in this place? never understood it.