cvat annotation view always using compressed images

[kristofbruyninckx]

When creating a task there are a few options to select image quality, chunk size and so on.

Setting the image quality to 100% does not matter for the annotation view, which always seems to compress the image. I also set tried setting the chunk size to 1. and disabling the force to use zip chunks as compressed data option

In every case, cvat downloads compressed images for the annotation view, which generates artifacts that are difficult to discern from actual defects for my specific dataset. By using chrome dev tools i can see that a chunk of data is downloaded from the api like this http://some_ip:8085/api/v1/tasks/10/data?type=chunk&number=0&quality=compressed. Manually changing compressed to original here gives me a zip with data in original quality. Is there any way for the annotation view to pass this parameter and use original quality? After all i chose 100% quality for a reason.

[nmanovic]

@kristofbruyninckx , I agree that having the switch in UI is a good idea. If you can contribute the feature, it will be great. I believe it is enough to have a switch inside player settings.

[jbohnslav]

Is there any progress on this issue? I find that even if I upload a task with --image_quality 100, the annotation view is very compressed.

[O-J1]

This is still an issue

[merlinwizzard]

fix all repositories

[jaroslavknotek]

Is there any workaround? I have images which are impossible to even see:

Input image(cropped):

Image in the editor:

I am not able to contribute in any way, however, I tried to "hard code" original compression in the code but it sees to be on ~30 different places. Is there any patch for this for the develop branch?