Closed GitEasonXu closed 5 years ago
I DON'T really convert darknet weights to tensorflow ckt, instead extract feature extractor weights darknet53.conv.74 format of darknet and save it as npz using save_npz.py
(modified from yad2k) This way, tensorflow weights can be initialized with them. I use train_net.py
to train the following neural network(head and detector), feature extractor is frozen at the time.
谢谢你的回复,
Github上有yolov3.weights to Keras模型的代码,但是一直找不到Tensorflow模型的实现,实在感谢。 如果我想要使用预训练模型进行再训练应该如何进行?还望指点。