search

jkjung-avt / yolov4_crowdhuman

A tutorial on training a DarkNet YOLOv4 model for the CrowdHuman dataset
MIT License
139 stars 49 forks source link

meaning of 608 in "Prepare_data.sh" execution #18

Closed ingbeeedd closed 3 years ago

ingbeeedd commented 3 years ago

When I download data, do specify 608 to build a directory name to make it easier to use later, regardless of the size of the data? So, the data doesn't change according to the arugument, does it? It automatically resize when I train the model anyway? Thank you

jkjung-avt commented 3 years ago

When preparing yolo (darknet) annotations, I filter out objects that are too small for the model to learn. Reference: https://github.com/jkjung-avt/yolov4_crowdhuman/issues/12#issuecomment-757396925

The filtering would be dependent upon input dimension. That's why I chose to create different data directories for models with different input dimensions.

ingbeeedd commented 3 years ago

Then, if you give 416x256 as input, what would be the change?

jkjung-avt commented 3 years ago

Then, if you give 416x256 as input, what would be the change?

It should be pretty straightforward.

  1. $ ./prepare_data.sh 416x256
  2. $ ./prepare_training.sh 416x256
  3. Create "cfg/yolov4-crowdhuman-416x256.cfg". Remember to re-calculate the "anchors" in the cfg.
  4. Train the model using the new cfg.