UNet2D.jl

This is a Julia wrapper for https://github.com/flavell-lab/unet2d

Instrunctions

Model creation and evaluation

path_weights = ...
model = create_model(1, 1, 32, path_weights=path_weights)

img_test_single = rand(Float64, 1024, 1024)
img_test_single = Float32.(standardize(img_test_single))

eval_model(img_test_single, model)

Batch evaluation

For batch evaluation, the input array dim should be (n, h, w):

img_test_batch = rand(Float64, 3, 1024, 1024)
img_test_batch = Float32.(standardize(img_test_batch))
eval_model(img_test_batch, model)

CPU evaluation

The package uses "cuda:0" as default device. To evaluate using CPU or anyother device, create a torch device:

device_cpu = UNet2D.py_torch.device("cpu")
model = create_model(1, 1, 32, device=device_cpu)
eval_model(img_test_single, model, device=device_cpu)