parameterstensorflowneural-networkdeep-learningimage-recognition

How many epochs does it take to train VGG-16

I'm training a VGG-16 model from scratch using a dataset containing 3k images. I use Tensorflow platform and 8 cpus without any gpu.

Training rate - 0.01,

Weight decay - 0.0005,

Momentum - 0.9,

Batch size - 64,

I've kept training for about three days. But the training accuracy has been unchanged, around 15%-20% after 20 epochs. Could anyone give me some hints to improve the accuracy?

Solution

  • It seems like I have used too large learning rate. Or weight decay does not work as it promises. After I changed my hyper parameters into,

    Training rate - 0.001,

    Weight decay - 0,

    Momentum - 0.9,

    Batch size - 64.

    Eveything is good. Now I can get accuracy around 90 percent after 25 epochs. Just for your information.