|
|
|
|
|
|
by tonic_section
2101 days ago
|
|
|
The model was trained on a fairly image (~1e6) dataset of diverse high-resolution natural images (the Openimages dataset) - so there was no particular training domain, and generalizes to images of arbitrary size/resolution/content well. There is a larger set of samples generated using the medium bitrate model which can be viewed in this Google Drive: https://drive.google.com/drive/folders/1lH1pTmekC1jL-gPi1fhE... One interesting failure model is that images dominated by high-frequency detail require a relatively large bitrate to store - see e.g. the last example in the Github README with the weird brickwork. Even though the model was trained to produce compressed representations with a soft constraint on the maximum bitrate, the filesize of the representation for this particular image is something like 60% above the nominal maximum. |
|
|