Open Wrestor opened 10 months ago
Is even possible to use gpu acceleration when resizing/hashing images(I am mainly talking about the available libraries)?
Duplicate videos are found by a separate library that extracts a certain number of images from the video stream and hashes them, so here, too, the main problem is the issue of resizing/hashing images and not hardware acceleration(I don't think hardware acceleration speeds up taking images from middle of video, but I could be wrong)
Is even possible to use gpu acceleration when resizing/hashing images(I am mainly talking about the available libraries)?
Duplicate videos are found by a separate library that extracts a certain number of images from the video stream and hashes them, so here, too, the main problem is the issue of resizing/hashing images and not hardware acceleration(I don't think hardware acceleration speeds up taking images from middle of video, but I could be wrong)
I see, I didn't know that, should I close this request then?
Is there anything more about it?
for reference with videos:
In the past I experimented with using gstreamer instead of ffmepg to decode videos. It is very easy to use GPU-decoding with gstreamer, however:
So I think GPU accelerated video decoding would be a bad idea
(FWIW the underlying library, vid_dup_finder_lib uses ffmpeg not gstreamer, but I am sure ffmpeg can use GPU decoding)
Feature Description Implement hardware acceleration to take advantage of specialized hardware and speed up czkawka even more. This would make finding similar photos and videos even more efficient. Similar to videoduplicatefinder