search

AlexeyAB / darknet

YOLOv4 / Scaled-YOLOv4 / YOLO - Neural Networks for Object Detection (Windows and Linux version of Darknet )
http://pjreddie.com/darknet/
Other
21.77k stars 7.96k forks source link

How to select cfg file in training? so many cfg files. #300

Open DanLiu0623 opened 6 years ago

DanLiu0623 commented 6 years ago

I use yolo-voc.cfg or tiny-yolo-voc.cfg generally due to my training data is voc format, But there are other cfg files, how should I choose ?

AlexeyAB commented 6 years ago

  1. These cfg files are for the detection - to say what objects and where are objects on the image: yolo-voc.cfg, yolo-voc.2.0.cfg, tiny-yolo-voc.cfg, yolo.cfg, yolo.2.0.cfg, tiny-yolo.cfg - they differ speed and accuracy: https://pjreddie.com/darknet/yolo/

  2. These cfg files are for the detection - to say what objects on the image: alexnet.cfg, cifar.cfg, darknet.cfg, vgg-16.cfg, darknet19.cfg, darknet448.cfg, resnet50.cfg, densenet201.cfg - they differ speed and accuracy: https://pjreddie.com/darknet/imagenet/

  3. There are recurrent netowkrs: rnn.cfg, rnn.train.cfg - to generate text in the style of great writers: https://pjreddie.com/darknet/rnns-in-darknet/

mechlya commented 4 years ago
  1. These cfg files are for the detection - to say what objects on the image: alexnet.cfg, cifar.cfg, darknet.cfg, vgg-16.cfg, darknet19.cfg, darknet448.cfg, resnet50.cfg, densenet201.cfg - they differ speed and accuracy: https://pjreddie.com/darknet/imagenet/

can I change this fils to train my own custom object detection with different architectures !

stephanecharette commented 4 years ago

You are replying to a very old message. I have a page that lists the various configuration files, and some information on each one: https://www.ccoderun.ca/darkmark/Configuration.html

And yes, you should take the one that matches what you'd like to do, and then modify it accordingly if you want to train your own custom neural network.