Open DaveHu-TVU opened 1 month ago
I have read the document https://www.intel.com/content/www/us/en/developer/articles/technical/support-va-api-vp9-k-svc-encoding-on-chromeos.html and this document which describes the basic process of encoding VP9 K-SVC with VAAPI, so I would like to find out more about the VAAPI VP9 K-SVC
VP9 SVC is handled by application , libva+media driver only provide support of DRC (dynamic resolution change) , it means , you could change the resolution in a sequence . it then provide the opportunity to enable SVC by spatial and temporal layer enabling. so, you could use different resolution for different layer in one sequence.
thanks for your reply.
you could change the resolution in a sequence -----> it means, there are different resolutions in a sequence? with one encoder pipeline?
yes, it means, smaller resolution picture could be reference of bigger resolution picture, change resolution without an IDR frame.
Thanks, does vpl or msdk have a sample with vp9 k-svc? and do you have more documents about VP9-KSVC? if K-SVC requires application handling, I need to know more details.
AFAIK, there are no sample in msdk or vpl. the link you pasted before include one 3x3 description. 3 layers spatial scaling and 3 temporal layers
Hi @XinfengZhang , I found the mfxExtVP9TemporalLayers in https://intel.github.io/libvpl/v2.10/API_ref/VPL_structs_encode.html#_CPPv4N23mfxExtVP9TemporalLayers5LayerE, it means, vpl had supported TemporalLayer and can't support SpatialLayer, right?
I downloaded the code of the latest libva and found that it doesn't search any field related to VP9 SVC. Does the latest libva version support VP9 SVC? thanks!