Continuing the Training for CS:GO

[MeekNasty]

I would like to continue the training you have done, with new images that it lacks, if you can point me in the right direction.

[FelipeVqs]

+1

[ofeksadlo]

It's based on yolov3-tiny model as shown in here: https://github.com/petercunha/Pine/tree/master/models You can continue the training as any darknet model can. First you'll need to create a decent dataset with this tutorial: https://www.youtube.com/watch?v=EGQyDla8JNU Then just follow along this tutorial: https://www.youtube.com/watch?v=mmj3nxGT2YQ But instead of using the pretrained backbone of the model (yolov4.conv.137) use this weights: https://github.com/petercunha/Pine/blob/master/models/yolov3-tiny.weights?raw=true And instead of the config he provide in the tutorial use this config: https://github.com/petercunha/Pine/blob/master/models/yolov3-tiny.cfg So when you want to begin the training your finel code should look like this: !./darknet detector train data/obj.data cfg/yolov3-tiny.cfg /mydrive/yolov4/backup/yolov3-tiny.weights -dont_show -map

For significantly change you will need a big dataset (At least 300 images).

[YADJ123]

if i want to train for a whole new game should i make my own data set and weights and start from zero or what? Also which obj should i be training for in a game like call of duty? Im guessing weapons, people and faces? Thanks