search

taipingeric / yolo-v4-tf.keras

A simple tf.keras implementation of YOLO v4
MIT License
140 stars 79 forks source link

Model Fails to load darknet weights when using custom classes. #5

Open nasheedyasin opened 3 years ago

nasheedyasin commented 3 years ago

Following is the message I recieve: failed to read all weights, # of unread weights: 0 load from ./darknet_base_model/yolov4.weights

I am training it on the ASL (American Sign Language) dataset and it has 26 classes from A-Z. The class file is formatted as per standard (1 class per line).

gajdosech2 commented 3 years ago

Is it necessary to use the pretrained weights in your case? The dataset you are using looks quite different from VOC/COCO anyway, so maybe you should just train the network from scratch.

nasheedyasin commented 3 years ago

They are different no doubt, but certain low level features that would be captured at the upper levels of the model would be fairly similar and I would ideally look to having those layers frozen, leaving the deeper layers, which more dataset specific features, unfrozen