If I only want to run the player identification (excluding jersey number recognition) and team affiliation modules to improve the overall model's speed, how can I disable the other modules or extract specific ones?

[JZ-Lay]

No description provided.

[victorjoos]

In the soccernet.yaml config :

sn-gamestate/sn_gamestate/configs/soccernet.yaml

Lines 30 to 38 in 17f9203

	- bbox_detector
	- reid
	- track
	- pitch
	- calibration
	- jersey_number_detect
	- tracklet_agg
	- team
	- team_side

You can remove the modules that you don't use, as long as they don't depend on earlier modules. If there is a module with a missing input (for example the reid module needs "bbox_ltwh" provided by the detector) you will get an error when the pipeline is being built

Hope this helps !

[JZ-Lay]

Thanks very much, I will have a try at once.

[JZ-Lay]

Excuse me, I have another question, I think repository of sn-gamestate seems overly redundant for my tasks. If I only want to run the model mentioned in the paper "Multi-task Learning for Joint Re-identification..." to achieve player identification and team affiliation tasks. Would it be possible to replicate this repository here for this purpose? Alternatively, could you provide me with some other suggestions (aside from replicating sn-gamestate)?