walkupup / spnet

Train

1. for first run, you can call gen_pattern() to generate DMD pattern. There is an already generated pattern "p1127.txt". Use it directly.

2. unzip "test_images.zip" from MNIST dataset.

3. use the following command to train a CNN on Mnist python main.py --datapath=data/test_images --batch_size 256 --epochs 10 --no_cuda True

Test

Run test.py
make sure the following 2 lines are consistent with the training code before testing:

1. model = SPNet('p1127.txt', 128)
2. model.load_state_dict(torch.load('output/params_14.pth'))