VMAF score of 100 coming for dolby vision 4K encoded content

[tarun-discovery]

I tried following version of command

ffmpeg -i source.mxf -i transcoded-dragons.mp4 -lavfi "[0:v]format=pix_fmts=yuv420p10le,setpts=PTS-STARTPTS[source];[1:v]scale=w=3840:h=2160:flags=bicubic[encoded];[encoded]setpts=PTS-STARTPTS[encoded1];[encoded1][source]libvmaf=model=path=./model/vmaf_4k_v0.6.1.json:feature='name=psnr|name=float_ssim|name=float_ms_ssim':n_threads=4:log_fmt=json:log_path=vmaf.json" -f null -

VMAF = 100

and

ffmpeg -loglevel debug -i source.mxf -i transcoded-dragons.mp4 -an -lavfi "[0:v]zscale=primaries=bt2020:transfer=smpte2084:matrix=bt2020nc,format=yuv420p10le,setpts=PTS-STARTPTS[source];[1:v]setpts=PTS-STARTPTS[dv_converted];[dv_converted]scale=w=3840:h=2160:flags=bicubic[encoded];[encoded][source]libvmaf=model=path=/content/model/vmaf_4k_v0.6.1.json:feature='name=psnr|name=float_ssim|name=float_ms_ssim|name=cambi':n_threads=4:log_fmt=json:log_path=abc.json" -f null -

VMAF = 99.7

These values clearly seems wrong,

Should we doing anything specific while calculating vmaf for dolby encoded HDR 4K content

[nilfm99]

Hi, I'd like to understand the issue better. Is the problem that the encoded content visually looks worse than the score of 100 is suggesting? If so, could you provide some information that could help us reproduce? Ideally the two input files themselves, but as an intermediate step, maybe seeing the elementary feature scores in vmaf.json could help us.