Sacha
Sacha
Yes, I saw that fp16 is in there now and forgot to close my PR. Thanks for heads up.
@Azkali is it possible to publish a new build on the repo with latest changes (especially the close stream memory leak)? Also, with rtsp enabled. Thanks. Edit: I'll build my...
This gives me: "NvMMLiteOpen : Block : BlockType = 261 NVMEDIA: Reading vendor.tegra.display-size : status: 6 NvMMLiteBlockCreate : Block : BlockType = 261 Assertion ((src_linesize) >= 0 ? (src_linesize) :...
Has anyone done this conversion yet? So I don't redo all their work.
Apparently nvidia has added hardware jetson decoder support to ffmpeg, 2 years ago, already. https://forums.developer.nvidia.com/t/hardware-accelerated-video-playback-with-l4t-ffmpeg/154019/7 There's also this fork, which has nvidia's method and this method (for encoding support mainly)...
I want to use FFMPEG of course, not l4t directly :).
This one works, #24 doesn't. I believe there's issues elsewhere where the width is being used instead of pitch. I ended up with images being cut off/green/corrupt with #24 With...
Definitely not above 3.5. I think 3.0.
Just a note: If you set the Travis CI to use the version of CMake you desire, then you never need to worry about compatibility -- it would fail to...
Getting 10fps without flip and 7fps with flip on a Titan V. Haven't been able to trace the model and preserve size. Also, torchvision op (deform conv2d) is an issue...