max_batches value - Githubissues

jkjung-avt / yolov4_crowdhuman

A tutorial on training a DarkNet YOLOv4 model for the CrowdHuman dataset

MIT License

140 stars 49 forks source link

max_batches value #30

Closed dlavrantonis closed 2 years ago

dlavrantonis commented 2 years ago

This is more a question than an issue. Since the crowdhuman608x608 dataset contains around 15000 photos for training, why is the max_batches set to 6000 only? In the instructions of yolov4 there is a guideline saying that the max_batches shall be at least equal to the number of your training images.

jkjung-avt commented 2 years ago

Refer to yolov4-crowdhuman-608x608.cfg:

batch=64 (stack 64 images as a batch/iteration during training)
max_batches=6000
So the total number of images used during training = 64 * 6,000 = 384,000

In fact, I determined the number 6,000 by looking at the "loss curve". Training the model longer probably doesn't help (lowering the loss) much.

dlavrantonis commented 2 years ago

thanks!