add resizing

infer.py

@@ -37,9 +37,9 @@ def infer_pipe(pipe, test_image, task_name, seed, device, video_depth=False):
         if max(test_image.shape[:2]) > 1024:
             # resize for a maximum size of 1024
             scale = 1024 / max(test_image.shape[:2])
-        elif min(test_image.shape[:2]) < 256:
-            # resize for a minimum size of 256
-            scale = 256 / min(test_image.shape[:2])
+        elif min(test_image.shape[:2]) < 384:
+            # resize for a minimum size of 384
+            scale = 384 / min(test_image.shape[:2])
         else:
             scale = 1.0
         new_shape = (int(test_image.shape[1] * scale), int(test_image.shape[0] * scale))