Old rules, first attach the official tutorial: http://caffe.berkeleyvision.org/gathered/examples/mnist.html
1. Necessary software
Because Caffe uses shell scripts that can only be run by Linux, first install wget (put wget into C:\windows\system32) and Git can run.
2. Then follow the official tutorial, first enter the root directory of the caffe path, and then open cmd and enter the command:
./data/mnist/get_mnist.sh
This command is to download the mnist data by opening the get_mnist.sh script in the /data/mnist directory. If there is an error in the cmd, you can directly enter and open the get_mnist.sh script. The effect is the same. After the operation is completed, the following 4 data files will appear:
Then continue to enter the following command, or enter the path to open the same
./examples/mnist/create_mnist.sh
If the file does not exist, you can create a create_mnist.sh by yourself. The specific code is as follows (Note: the ninth line of BUILD may have a different path to the old version, and modify it according to your own path):
#!/usr/bin/env sh # This script converts the mnist data into lmdb/leveldb format, # depending on the value assigned to $BACKEND. set -e EXAMPLE=. DATA = .. / .. / data / mnist BUILD=../../scripts/build/examples/mnist/Release BACKEND="lmdb" echo "Creating ${BACKEND}..." rm -rf $EXAMPLE/mnist_train_${BACKEND} rm -rf $EXAMPLE/mnist_test_${BACKEND} $BUILD/convert_mnist_data.exe $DATA/train-images-idx3-ubyte \ $DATA/train-labels-idx1-ubyte $EXAMPLE/mnist_train_${BACKEND} --backend=${BACKEND} $BUILD/convert_mnist_data.exe $DATA/t10k-images-idx3-ubyte \ $DATA/t10k-labels-idx1-ubyte $EXAMPLE/mnist_test_${BACKEND} --backend=${BACKEND} echo "Done." read -p "Enter to continue..."
After the run is complete, two folders, mnist_test_lmdb and mnist_train_lmdb, will appear:
cmd shows:
3. Open lenet_solver.prototxt under the path /scripts/build/examples/mnist/Release (the paths of different versions of caffe are different, and the paths of some old versions of caffe are: /Build/x64/Release), according to your own situation Change parameters:
The second line: If lenet_train_test.prototxt and lenet_solver.prototxt are not in the same path, you need to write the path where lenet_train_test.prototxt is located before it
Line 23: snapshot_prefix: the generated model is the generated training model, you can modify the path according to yourself
The last line selects whether the installed caffe is CPU or GPU. I installed the GPU version here.
Note: Do not copy the file path directly, because the path separator in it is / instead of \. If you use \ and run it later, the following error will occur (the same is true for the following steps, if you are not sure, just follow my writing ):
The result after modifying the parameters:
# The train/test net protocol buffer definition net: "lenet_train_test.prototxt" # test_iter specifies how many forward passes the test should carry out. # In the case of MNIST, we have test batch size 100 and 100 test iterations, # covering the full 10,000 testing images. test_iter: 100 # Carry out testing every 500 training iterations. test_interval: 500 # The base learning rate, momentum and the weight decay of the network. base_lr: 0.01 momentum: 0.9 weight_decay: 0.0005 # The learning rate policy lr_policy: "inv" gamma: 0.0001 power: 0.75 # Display every 100 iterations display: 100 # The maximum number of iterations max_iter: 10000 # snapshot intermediate results snapshot: 5000 snapshot_prefix: "E:/CaffeSource/caffe/data/mnist/model" # solver mode: CPU or GPU solver_mode: GPU
4. Open lenet_train_test.prototxt (the file in the second line above)
Then change the path in the figure above. These two files are the files downloaded when the ./data/mnist/get_mnist.sh command is executed, and their paths are added.
5. Create a new train_lenet.txt file in the directory \examples\mnist, add the following paragraph, and then change the suffix to .bat
..\..\Build\x64\Release\caffe.exe train --solver="lenet_solver.prototxt" --gpu 0 pause
Or modify the train_lenet.sh file in this directory:
#!/usr/bin/env sh set -e BUILD=../../Build/x64/Release/ echo "Training lenet_solver.prototxt..." $BUILD/caffe.exe train --solver=lenet_solver.prototxt $@ echo "Done." read -p "Enter to continue..."
6. Run the file, and the result after running for a few minutes is as follows:
If no error is reported, the test is a success!
It can be seen that the accuracy is 99%, the trained model is saved in lenet_iter_10000.caffemodel, and the training state is saved in lenet_iter_10000.solverstate. The results are as follows: