Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
The authors evaluate JEVQA, an approach that repurposes a general-purpose commercial decision model (Jev) for zero-shot video quality assessment across AV1, H.264, HEVC, and VP9 encodes using metadata, bitstream, and pixel inputs. This investigates whether foundation-style decision systems can bypass the rigid retraining cycles required by standard, codec-specific instrumental video quality metrics. Remarkably, using encoding metadata alone, the zero-shot model achieves Pearson correlations on par with the standardized ITU-T P.1204.1 metric ($r = 0.737$ vs. $0.733$ against VMAF, and $r = 0.879$ vs. $0.898$ against subjective MOS), climbing to $r = 0.824$ when combined with bitstream and pixel features.
Raw encoding metadata fed into an off-the-shelf decision model matches dedicated ITU-T video quality benchmarks without a single gradient step of task-specific training.
Instrumental quality models for video quality prediction are usually trained for a fixed set of codecs or other input features, and every new input variable requires retraining. Novel, general-purpose decision models can answer questions without task-specific training, but it is unclear whether they can judge video quality. We evaluate Jev, a commercial ``System One''model that returns probability distributions over a provided answer scale, as a zero-shot video quality model. We call the resulting method JEVQA. In a first study on 1,936 AV1, H.264, HEVC, and VP9 encodes of 22 sources, scored against VMAF as ground truth, using encoding metadata only, JEVQA reached a Pearson correlation of 0.737, on par with the standardized ITU-T P.1204.1 model (0.733). Giving the model bitstream data raised the accuracy to 0.797, and combined pixel-based and bitstream features raised it to 0.824. A pixel-only variant failed in our tests. In a second study, using H.264, HEVC, and VP9 encodes in the AVT-VQDB-UHD-1 database, the metadata-only model reached a correlation of 0.879 with MOS, close to P.1204.1 (0.898). Bitstream statistics did not help there. Our results show that trained models on the same features remain clearly ahead in both studies, but that zero-shot classifiers are promising.