diff --git a/vision_agent/tools/tools.py b/vision_agent/tools/tools.py index fd315bed..fa06a823 100644 --- a/vision_agent/tools/tools.py +++ b/vision_agent/tools/tools.py @@ -588,7 +588,9 @@ def __call__(self, image: Union[str, ImageType], prompt: str) -> Dict: "prompt": prompt, "tool": "few_shot_counting", } - return _send_inference_request(data, "tools") + resp_data = _send_inference_request(data, "tools") + resp_data["heat_map"] = np.array(b64_to_pil(resp_data["heat_map"][0])) + return resp_data class VisualQuestionAnswering(Tool):