VMAF Re-implementation on PyTorch: Some Experimental Results
Fuente:
arXiv
Salvato in:
| Autori principali: | , |
|---|---|
| Natura: | Preprint |
| Pubblicazione: |
2023
|
| Soggetti: | |
| Accesso online: | |
| Tags: |
Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
|
| _version_ | 1866916324972691456 |
|---|---|
| author | Aistov, Kirill Koroteev, Maxim |
| author_facet | Aistov, Kirill Koroteev, Maxim |
| contents | Based on the standard VMAF implementation we propose an implementation of VMAF using PyTorch framework. For this implementation comparisons with the standard (libvmaf) show the discrepancy $\lesssim 10^{-2}$ in VMAF units. We investigate gradients computation when using VMAF as an objective function and demonstrate that training using this function does not result in ill-behaving gradients. The implementation is then used to train a preprocessing filter. It is demonstrated that its performance is superior to the unsharp masking filter. The resulting filter is also easy for implementation and can be applied in video processing tasks for video copression improvement. This is confirmed by the results of numerical experiments. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2310_15578 |
| institution | arXiv |
| publishDate | 2023 |
| record_format | arxiv |
| spellingShingle | VMAF Re-implementation on PyTorch: Some Experimental Results Aistov, Kirill Koroteev, Maxim Machine Learning Computer Vision and Pattern Recognition Based on the standard VMAF implementation we propose an implementation of VMAF using PyTorch framework. For this implementation comparisons with the standard (libvmaf) show the discrepancy $\lesssim 10^{-2}$ in VMAF units. We investigate gradients computation when using VMAF as an objective function and demonstrate that training using this function does not result in ill-behaving gradients. The implementation is then used to train a preprocessing filter. It is demonstrated that its performance is superior to the unsharp masking filter. The resulting filter is also easy for implementation and can be applied in video processing tasks for video copression improvement. This is confirmed by the results of numerical experiments. |
| title | VMAF Re-implementation on PyTorch: Some Experimental Results |
| topic | Machine Learning Computer Vision and Pattern Recognition |
| url | https://arxiv.org/abs/2310.15578 |