Hello！When I followrd the page to deal with data at step 2, I got a problem:
command:ffmpeg -pix_fmt yuv420p -s 960x512 -i ./yuv/001.yuv -vf scale=240x128:flags=bicubic ./downsample/001_down4.yuv
problem:
Stream mapping:
Stream #0:0 -> #0:0 (rawvideo (native) -> rawvideo (native))
Press [q] to stop, [?] for help
Output #0, rawvideo, to './downsample/001_down4.yuv':
Metadata:
encoder : Lavf58.45.100
Stream #0:0: Video: rawvideo (I420 / 0x30323449), yuv420p, 240x128 [SAR 1:1 DAR 15:8], q=2-31, 9216 kb/s, 25 fps, 25 tbn, 25 tbc
Metadata:
encoder : Lavc58.91.100 rawvideo
[rawvideo @ 0x5e19340] Packet corrupt (stream = 0, dts = 13).
./yuv/001.yuv: corrupt input packet in stream 0
[rawvideo @ 0x5e1d000] Invalid buffer size, packet size 460800 < expected frame_size 737280
Error while decoding stream #0:0: Invalid argument
frame= 13 fps=0.0 q=-0.0 Lsize= 585kB time=00:00:00.52 bitrate=9216.0kbits/s speed=18.9x
video:585kB audio:0kB subtitle:0kB other streams:0kB global headers:0kB muxing overhead: 0.000000%
So, how can I solve it?
Posted by: silence @ Feb. 23, 2022, 8:22 a.m.It seems that the error was caused by the wrong resolution of the input video, i.e., the /yuv/001.yuv is not 960x512, but the command tries to read it as this resolution. Could you please make sure that the input video comes from "ffmpeg -i 001.mkv -pix_fmt yuv420p 001.yuv", and 001.mkv is the one in the link for Track 3?
If there is still problem, feel free to write to me at ren.yang@vision.ee.ethz.ch
Posted by: RenYang @ Feb. 23, 2022, 8:57 a.m.I downloaded the data from the following link at the website https://codalab.lisn.upsaclay.fr/competitions/1497#participate:
------------------------------------------------------------link---------------------------------------------------------------------------------
Data access
Training data:
Raw Video: https://data.vision.ee.ethz.ch/reyang/NTIRE2022/train/train_down4_gt.zip
Compressed Video: https://data.vision.ee.ethz.ch/reyang/NTIRE2022/train/train_down4_QP37.zip
Video information: https://data.vision.ee.ethz.ch/reyang/NTIRE2022/train/data_train_3.xlsx
---------------------------------------------------------------------------------------------------------------------------------------------
Is that mean I just need perform step 5 for each link (Raw Video and Compressed Video) ?
Step 1 - 4 are the description for how to get the compressed video, because you could also collect and make more training data in the same way.
Step 5 is to clarify that we calculate the results in RGB domain given those png files (compressed, enhanced, raw).
Actually, you could deal with data in either compressed domain (.mkv, directly download), YUV domain (.yuv, convert to YUV by Step 1) or RGB domain (.png, convert to PNG by Step 5). The Step 5 only means that we will evaluate the performance in RGB domain, no matter how you deal with the data.
Posted by: RenYang @ Feb. 23, 2022, 9:17 a.m.I got it, thanks.
Posted by: silence @ Feb. 23, 2022, 9:19 a.m.If you want to perform Step 2, then make sure that the YUV file, e.g., 001.yuv, comes from 001.mkv in the raw videos link. If there is still problem, please send me the mkv file that you got the error by email, and I will look into it
Posted by: RenYang @ Feb. 23, 2022, 9:20 a.m.