Precious and recall are 0

WongKinYiu / yolor

implementation of paper - You Only Learn One Representation: Unified Network for Multiple Tasks (https://arxiv.org/abs/2105.04206)

GNU General Public License v3.0

1.98k stars 524 forks source link

Precious and recall are 0 #240

Open GreenHandForCoding opened 2 years ago

GreenHandForCoding commented 2 years ago

使用的数据集是我在yolov5使用过的，在yolor中训练时很多参数都为0

Swearinbag commented 2 years ago

Can you give more info? It could take some time before the model actually starts learning. Max I've had was around 10 epochs before it started.

JYSKKwithm commented 2 years ago

I also have this problem.

nhatanh174 commented 2 years ago

You need to change "class" and "filter" parameter in cfg file. filter = (number of class + 5)*3

Swearinbag commented 2 years ago

What @nhatanh174 said is indeed very important- but it could also be that your model just has a lot to learn and can take some time to do this. wait for 20 epochs and if nothing changes with the precision and recall you could look for other issues :) Also be sure that input size is somewhat near the "img_size" given to the model. I've observed that a lot of stretching/rescaling is very non beneficial to the training performance.