MMaxV
commented
6 years ago This file, actually, is a description of the model that you are using. It contains all the layers with it's parameters. In order to your dataset you should change these parameters(for example yolov3.cfg):
Line 3: set batch=24, this means we will be using 24 images for every training step
Line 4: set subdivisions=8, the batch will be divided by 8 to decrease GPU VRAM requirements.
Line 603: set filters=(classes + 5)3 in our case filters=21
Line 610: set classes=2, the number of categories we want to detect
Line 689: set filters=(classes + 5)3 in our case filters=21
Line 696: set classes=2, the number of categories we want to detect
Line 776: set filters=(classes + 5)*3 in our case filters=21
Line 783: set classes=2, the number of categories we want to detect.
(classes are different objects in your dataset, for example cats, dogs, ...)
It can be differ in other cfg files.
You can use any of cfg files, but you need to change those parameters. Also remember to change your .name and .data files.
han-so1omon
YunYang1994
FranciscoReveriano
ghost
zaimpower
I'm trying to train a classifier model using darknet, and I followed this tutorial, and my training command is like this:
./darknet classifier train cfg/my_classifier.data cfg/darknet19.cfgthe 'my_classifier.data' file has been modified according to my dataset. Anyhow, I'm a little confused about the darknet19.cfg. My question is that there are dozen of cfg files the darknet/cfg folder, can I just choose one cfg file and use it to train a classifier model, or I can just use certain cfg file to train it? As far as I know, the darknet19_448.cfg can be used during detector-training process, and it seems there's no big difference between darknet19.cfg and darknet19_448.cfg. Could someone please explain for me. FYI, I've trained my classifier model for about 10000 times, and it worked out not very well, it may be my dataset's problem, but I want to be sure, that's why I want to know if I can use another cfg file to achieve better accuracy.