Unclear inference results.

[LoaferCH]

After running inference.py, output is saved as .png images into logs folder. However it is unclear what is the original data.
It could be better if forward() method returned a tuple of both outs and batch

innofw/innofw/core/models/torch/lightning_modules/segmentation.py

Lines 64 to 68 in f0bfa99

	def forward(self, batch: torch.Tensor) -> torch.Tensor:
	"""Make a prediction"""
	logits = self.model(batch)
	outs = (logits > self.threshold).to(torch.uint8)
	return outs

So both original image and predict for it could be saved with cv2.imwrite at:

innofw/innofw/core/datamodules/lightning_datamodules/segmentation_hdf5_dm.py

Lines 98 to 108 in f0bfa99

	def save_preds(self, preds, stage: Stages, dst_path: pathlib.Path):
	out_file_path = dst_path / "results"
	os.mkdir(out_file_path)
	for preds_batch in preds:
	for i , pred in enumerate(preds_batch):
	pred = pred.numpy()
	pred[pred < 0.3] = 0
	pred[pred > 0] = 255
	filename = out_file_path / f"out_{i}.png"
	cv2.imwrite(filename, pred[0])
	logging.info(f"Saved result to: {out_file_path}")

Originals could be saved with original_ prefix in the same folder as preds.