There is no "parser.add_argument("--detection") in eval_inference_vidhoi.py

nizhf / hoi-prediction-gaze-transformer

MIT License

22 stars 3 forks source link

There is no "parser.add_argument("--detection") in eval_inference_vidhoi.py #5

Closed ThePassedWind closed 1 year ago

ThePassedWind commented 1 year ago

I'd like to train a model of Detection Mode, but your training code is maybe only for Oracle Mode.

nizhf commented 1 year ago

We always train the model first with ground-truth bbox, then test with detected bbox. We do not have the ground-truth interaction for detected human-object pairs.

ThePassedWind commented 1 year ago

Got it