Evaluation Script Fails Due to Missing or Malformed JSON Output

[StanislawOlech]

We continue to encounter errors related to the evaluation script when running in two different environments. The traceback is as follows:

File "scripts/eval.py", line 86, in
dataset_list = [trackeval.datasets.MouseSIS(dataset_config)]
File "path\MouseSIS.\src\third_party\TrackEval\trackeval\datasets\mouse_sis.py", line 106, in init
raise TrackEvalException(tracker_dir_path + ' does not contain exactly one json file.')
TypeError: unsupported operand type(s) for

While the immediate error is caused by a TypeError—specifically, trying to concatenate a string with a non-string type when raising a TrackEvalException—this is merely a symptom of a deeper problem.
The root cause appears to be that inference.py does not generate any .json file in the output directory, which the evaluation script expects. Instead, it produces a .yaml file. Even when this YAML file is manually converted to JSON, it likely fails to conform to the expected schema required by the evaluation tools, particularly hota.py, which leads to additional downstream issue