Merge pull request #69 from GeorgeBatch/master

minor fixes: (1) tree fusion vs concat (2) sort magnifications after parsing: 0 2 vs 2 0 (3) remove optimizer from test()

compute_feats.py

@@ -81,7 +81,7 @@ def compute_feats(args, bags_list, i_classifier, save_path=None, magnification='
             os.makedirs(os.path.join(save_path, bags_list[i].split(os.path.sep)[-2]), exist_ok=True)
             df.to_csv(os.path.join(save_path, bags_list[i].split(os.path.sep)[-2], bags_list[i].split(os.path.sep)[-1]+'.csv'), index=False, float_format='%.4f')
 
-def compute_tree_feats(args, bags_list, embedder_low, embedder_high, save_path=None, fusion='fusion'):
+def compute_tree_feats(args, bags_list, embedder_low, embedder_high, save_path=None):
     embedder_low.eval()
     embedder_high.eval()
     num_bags = len(bags_list)
@@ -107,10 +107,14 @@ def compute_tree_feats(args, bags_list, embedder_low, embedder_high, save_path=N
                         img = Image.open(high_patch)
                         img = VF.to_tensor(img).float().cuda()
                         feats, classes = embedder_high(img[None, :])
-                        if fusion == 'fusion':
+
+                        if args.tree_fusion == 'fusion':
                             feats = feats.cpu().numpy()+0.25*feats_list[idx]
-                        if fusion == 'cat':
+                        elif args.tree_fusion == 'cat':
                             feats = np.concatenate((feats.cpu().numpy(), feats_list[idx][None, :]), axis=-1)
+                        else:
+                            raise NotImplementedError(f"{args.tree_fusion} is not an excepted option for --tree_fusion. This argument accepts 2 options: 'fusion' and 'cat'.")
+
                         feats_tree_list.extend(feats)
                 sys.stdout.write('\r Computed: {}/{} -- {}/{}'.format(i+1, num_bags, idx+1, len(low_patches)))
             if len(feats_tree_list) == 0:
@@ -133,6 +137,7 @@ def main():
     parser.add_argument('--weights', default=None, type=str, help='Folder of the pretrained weights, simclr/runs/*')
     parser.add_argument('--weights_high', default=None, type=str, help='Folder of the pretrained weights of high magnification, FOLDER < `simclr/runs/[FOLDER]`')
     parser.add_argument('--weights_low', default=None, type=str, help='Folder of the pretrained weights of low magnification, FOLDER <`simclr/runs/[FOLDER]`')
+    parser.add_argument('--tree_fusion', default='cat', type=str, help='Fusion method for high and low mag features in a tree method [cat|fusion]')
     parser.add_argument('--dataset', default='TCGA-lung-single', type=str, help='Dataset folder name [TCGA-lung-single]')
     args = parser.parse_args()
     gpu_ids = tuple(args.gpu_index)
@@ -238,7 +243,7 @@ def main():
     bags_list = glob.glob(bags_path)
 
     if args.magnification == 'tree':
-        compute_tree_feats(args, bags_list, i_classifier_l, i_classifier_h, feats_path, 'cat')
+        compute_tree_feats(args, bags_list, i_classifier_l, i_classifier_h, feats_path)
     else:
         compute_feats(args, bags_list, i_classifier, feats_path, args.magnification)
     n_classes = glob.glob(os.path.join('datasets', args.dataset, '*'+os.path.sep))

deepzoom_tiler.py

@@ -254,7 +254,7 @@ def nested_patches(img_slide, out_base, level=(0,), ext='jpeg'):
     parser.add_argument('-o', '--objective', type=float, default=20, help='The default objective power if metadata does not present [20]')
     parser.add_argument('-t', '--background_t', type=int, default=15, help='Threshold for filtering background [15]')  
     args = parser.parse_args()
-    levels = tuple(args.magnifications)
+    levels = tuple(sorted(args.magnifications))
     assert len(levels)<=2, 'Only 1 or 2 magnifications are supported!'
     path_base = os.path.join('WSI', args.dataset)
     if len(levels) == 2:

train_tcga.py

@@ -62,7 +62,7 @@ def dropout_patches(feats, p):
     sampled_feats = np.concatenate((sampled_feats, pad_feats), axis=0)
     return sampled_feats
 
-def test(test_df, milnet, criterion, optimizer, args):
+def test(test_df, milnet, criterion, args):
     milnet.eval()
     csvs = shuffle(test_df).reset_index(drop=True)
     total_loss = 0
@@ -188,8 +188,8 @@ def main():
         train_path = shuffle(train_path).reset_index(drop=True)
         test_path = shuffle(test_path).reset_index(drop=True)
         train_loss_bag = train(train_path, milnet, criterion, optimizer, args) # iterate all bags
-        test_loss_bag, avg_score, aucs, thresholds_optimal = test(test_path, milnet, criterion, optimizer, args)
-        if args.dataset=='TCGA-lung':
+        test_loss_bag, avg_score, aucs, thresholds_optimal = test(test_path, milnet, criterion, args)
+        if args.dataset.startswith('TCGA-lung'):
             print('\r Epoch [%d/%d] train loss: %.4f test loss: %.4f, average score: %.4f, auc_LUAD: %.4f, auc_LUSC: %.4f' % 
                   (epoch, args.num_epochs, train_loss_bag, test_loss_bag, avg_score, aucs[0], aucs[1]))
         else:
@@ -201,7 +201,7 @@ def main():
             best_score = current_score
             save_name = os.path.join(save_path, str(run+1)+'.pth')
             torch.save(milnet.state_dict(), save_name)
-            if args.dataset=='TCGA-lung':
+            if args.dataset.startswith('TCGA-lung'):
                 print('Best model saved at: ' + save_name + ' Best thresholds: LUAD %.4f, LUSC %.4f' % (thresholds_optimal[0], thresholds_optimal[1]))
             else:
                 print('Best model saved at: ' + save_name)