Minor changes to evaluation script

eval.py

@@ -139,6 +139,7 @@ def evaluate(
         classes=None,
         colors=None
     ):
+        metric = MeanAveragePrecision(class_metrics=args['verbose'])
         n_threads = torch.get_num_threads()
         # FIXME remove this and make paste_masks_in_image run on the GPU
         torch.set_num_threads(1)
@@ -172,13 +173,11 @@ def evaluate(
                 preds.append(preds_dict)
                 target.append(true_dict)
             #####################################
-
             outputs = [{k: v.to(cpu_device) for k, v in t.items()} for t in outputs]
 
         # gather the stats from all processes
         metric_logger.synchronize_between_processes()
         torch.set_num_threads(n_threads)
-        metric = MeanAveragePrecision(class_metrics=args['verbose'])
         metric.update(preds, target)
         metric_summary = metric.compute()
         return metric_summary

utils/validate.py

@@ -78,16 +78,18 @@
     # Model configurations
     IMAGE_WIDTH = args['img_size']
     IMAGE_HEIGHT = args['img_size']
-
     valid_dataset = create_valid_dataset(
-        VALID_DIR_IMAGES, VALID_DIR_LABELS, 
-        IMAGE_WIDTH, IMAGE_HEIGHT, CLASSES
-    )
+            VALID_DIR_IMAGES, 
+            VALID_DIR_LABELS, 
+            args['img_size'],
+            CLASSES,
+            square_training=True
+        )
     valid_loader = create_valid_loader(valid_dataset, BATCH_SIZE, NUM_WORKERS)
 
     coco_evaluator, stats = evaluate(
             model, 
             valid_loader, 
             device=DEVICE,
-            classes=CLASSES,
-    )
+            # classes=CLASSES,
+    )