Opened 9 months ago

#1610 new defect

Frame lost when the default NNPF config files are applied in coding.

Reported by: shurunwang Owned by:
Priority: minor Milestone:
Component: VTM Version: VTM-21.2
Keywords: Cc: ksuehring, XiangLi, fbossen, jvet@…

Description

VTM version is VTM-21.2, and the encoding configuration is random access (RA) with the given random access config file.
Using the default NNPF config files to compress video, frames could be lost in the decoded YUV file in the decoding process, while no frame lost in the reconstructed YUV file in encoding process.
Here are the sequence names and the related settings when frames are lost. (QP is 63 to save coding time.)

sequence name FrameSkip FrameToBeEncoded BitDepth
BQMall_832x480_60 471 129 8
BQSquare_416x240_60 471 129 8
BasketballDrill_832x480_50 0 500 8
PartyScene_832x480_50 0 500 8
BasketballPass_416x240_50 0 500 8
BlowingBubbles_416x240_50 0 500 8
TVD-01_1 1500 500 8
TVD-01_2 2000 500 8
TVD-01_3 2500 500 8
TVD-02_1 0 636 10
TVD-03_1 0 500 10
TVD-03_2 500 500 10
TVD-03_3 1000 500 10

Note: The TVD* sequences can be downloaded from https://multimedia.tencent.com/resources/tvd, clicking "Download Video Sequences with Annotations for Object Tracking" button and following the illustrations in https://jvet-experts.org/doc_end_user/documents/31_Geneva/wg11/JVET-AE2031-v1.zip to convert the .mp4 files into YUV files.


Change history (0)

Note: See TracTickets for help on using tickets.