fix bbox coords outside the image, countgd return types

vision_agent/tools/tool_utils.py

@@ -1,7 +1,7 @@
 import inspect
 import logging
 import os
-from typing import Any, Callable, Dict, List, MutableMapping, Optional, Tuple
+from typing import Any, Callable, Dict, List, MutableMapping, Optional, Tuple, Union
 
 import pandas as pd
 from IPython.display import display
@@ -34,7 +34,7 @@ def send_inference_request(
  files: Optional[List[Tuple[Any, ...]]] = None,
  v2: bool = False,
  metadata_payload: Optional[Dict[str, Any]] = None,
-) -> Dict[str, Any]:
+) -> Union[Dict[str, Any], List[Dict[str, Any]]]:
  # TODO: runtime_tag and function_name should be metadata_payload and now included
  # in the service payload
  try:

vision_agent/tools/tools.py

@@ -1660,7 +1660,10 @@ def overlay_counting_results(
  fontsize = max(10, int(min(width, height) / 80))
  pil_image = ImageEnhance.Brightness(pil_image).enhance(0.5)
  draw = ImageDraw.Draw(pil_image)
- font = ImageFont.load_default(size=fontsize)
+ font = ImageFont.truetype(
+ str(resources.files("vision_agent.fonts").joinpath("default_font_ch_en.ttf")),
+ fontsize,
+ )
 
  for i, elt in enumerate(instances):
  label = f"{i}"

vision_agent/utils/image_utils.py

@@ -181,7 +181,7 @@ def denormalize_bbox(
  raise ValueError("Bounding box must be of length 4.")
 
  arr = np.array(bbox)
- if np.all((arr >= 0) & (arr <= 1)):
+ if np.all((arr[:2] >= 0) & (arr[:2] <= 1)):
  x1, y1, x2, y2 = bbox
  x1 = round(x1 * image_size[1])
  y1 = round(y1 * image_size[0])