training result questions

[sobikim]

Hi.

I have some questions about training results.

1. In the figure below, does "t=0.20" mean the threshold for binarization of voxels?
   

2. When training, there are a lot of images in the "./output/images/time/test" folder. What are these images? It doesn't look like a prediction, is it the ground truth?

Thank you!

[hzxie]

1. Yes
2. ./output is in the .gitignore. If there are files in the ./output folder, they must be generated by yourself.

[sobikim]

Maybe I didn't ask clearly.

When I train Pix2Vox, a lot of images of voxels are created in that folder("./output/images/2022-06-30T04:43:11.767915/test").

Is this a predicted voxel? or ground truth voxel?

Thank you!

[hzxie]

The images are generated with the following code.

Pix2Vox/core/test.py

Lines 150 to 156 in f1b8282

	rendering_views = utils.binvox_visualization.get_volume_views(gv, os.path.join(img_dir, 'test'),
	epoch_idx)
	test_writer.add_image('Test Sample#%02d/Volume Reconstructed' % sample_idx, rendering_views, epoch_idx)
	gtv = ground_truth_volume.cpu().numpy()
	rendering_views = utils.binvox_visualization.get_volume_views(gtv, os.path.join(img_dir, 'test'),
	epoch_idx)
	test_writer.add_image('Test Sample#%02d/Volume GroundTruth' % sample_idx, rendering_views, epoch_idx)

The images of prediction volumes are overwritten by the GT volumes. Therefore, the images are GT volumes.
These results are used for visualization in Tensorboard.

[sobikim]

I understand.
Thank you always!!