Sample Test Loader v1

Supported frameworks: torch.

Arguments

Similar arguments to the Sample Training Loader, except that augmentation, shuffling, and batching are not necessary at test time.

Key	Description	Default	Type
samples	Number of samples to load per pixel	8	int
num_workers	Number of data loading processes	4	int
framework	Deep learning framework to use	torch	string

The buffers argument is the same as for the Sample Training Loader.

Usage

The test loader makes it easy to run inference on multiple test sequences and save the inferred images. You can use these functions as in the following example:

test_set = hydra.utils.instantiate(cfg['test_data'])
# or
test_set = Noisebase('sample_test8_v1')

test_set.save_dir(output_folder) # Where to save the inferred images
for sequence in test_set:
    first = True
    for i, frame in enumerate(sequence.frames):
        frame = sequence.to_torch(frame, model.device) # Convenience function to convert frames

        # Temporal initialization for NPPD in this example
        if first:
            first = False
            model.temporal = model.temporal_init(frame)

        with torch.no_grad(): # Save memory at test time
            output = model.test_step(frame)

        sequence.save(i, output) # Save inferred images asynchronously
    sequence.join() # Wait for images to finish saving in the background

The frame value here is compatible with batches produced by the Sample Training Loader.