Caffe fine tuning of the generated model, testing (Linux)

Thank Bo Lynnandwei, wrote the blog. Or because of its inspiration, so I thought of how to fine-tune the model, hey, vegetables than hurt.

It's still under the CIFAR10 model. You want to fine-tune the models that have been generated. Enter commands below the Caffe-master folder

./build/tools/caffe.bin Train-model=examples/cifar10/cifar10_quick_train_test.prototxt

-solver=examples/cifar10/cifar10_quick_solver.prototxt

-weight=examples/cifar10/cifar10_quick_iter_5000.caffemodel-gpu=0

The model here is the structure of this network. Configuration is the configuration file, and the weights are the initialization weights. Since the configuration file is directly inside the 4000 save a model. So it's kind of funny. 4000 of that model has changed. The 5000 model is still intact.

Caffe fine tuning of the generated model, testing (Linux)