Opened 12 years ago Closed 12 years ago #795 closed defect (fixed)When FramesTobeEncoded > actual num frames, encoder report is different
Description
I encoded the 150-frame Traffic sequence, but mistakenly specified 300 frames as FramesToBeEncoded. The encoder ran fine, but the reported rate and maybe PSNR averages were not the same as when I specified FramesToBeEncoded to be 150. I didn't yet repeat the test so someone may want to doublecheck this on a short sequence to see if this is a bug. Change History (7)comment:1 Changed 12 years ago by DefaultCC Plugin
comment:2 Changed 12 years ago by ksuehringcomment:3 Changed 12 years ago by cohen
I believe it was all-intra main configuration. comment:4 Changed 12 years ago by cohen
I just did a quick test on RaceHorses (Class D), AI-Main configuration, HM8.0, specifying the correct number of frames to be encoded (300), and then an incorrect number (600). The rates/PSNRs are different:
FramesToBeEncoded: 300
FramesToBeEncoded: 600 comment:5 Changed 12 years ago by davidf
The cause is likely due to the file reader failing to spot it hasn't loaded the frame. Can you check the log ans see if it encodes an extra picture? comment:6 Changed 12 years ago by cohen
Yes, that seems to be the case, and the bits/PSNRs reported for this extra picture are exactly the same as the real last picture. Here's a diff for the good and bad encoder outputs, followed by the relevant lines of the bad encoder output file:
diff good.txt bad.txt | more
329a330
334c335
339c340
352c353
354c355
Encoder output file for bad run:
POC 298 TId: 0 ( I-SLICE, nQP 37 QP 37 ) 23328 bits [Y 32.7065 dB U 35.5228 dB V 37.4423 dB] [ET 1 ] [L0 ] [L1 ] [MD5:3d512c213370e0593443d5ecf6ddc675,f29b3\ comment:7 Changed 12 years ago by davidf
Fixed in r2918. Note: See
TracTickets for help on using
tickets. | This list contains all users that will be notified about changes made to this ticket. These roles will be notified: Reporter, Owner, Subscriber, Participant
|
Was that a random-access GOP? My first guess is that the size of the last GOP was not adapted when the last frame was encountered.