Netflix开源Emmy获奖视频质量评估工具VMAF

Netflix开源了VMAF（感知视频质量评估算法），这是一款获得Emmy奖的视频质量评估工具1。该项目包含独立的C库libvmaf及其Python封装库1，为用户提供了多种灵活的使用方式。

VMAF工具集集成了FFmpeg、提供Docker镜像支持，并可在Windows平台上构建1。该工具包含PSNR、SSIM、MS-SSIM等多种视频质量指标1。在发展历程中，Netflix于2020年12月发布libvmaf v2.0.0版本，性能提升了2倍1；2023年12月推出libvmaf v3.0.0版本1；2026年6月发布了新的VMAF v1模型集合1。此外，Netflix还在2021年10月开源了CAMBI条纹检测工具1。

Netflix has open-sourced VMAF, an Emmy-winning algorithm for assessing perceptual video quality 1. The toolkit includes a standalone C library called libvmaf alongside its Python wrapper, offering multiple integration options such as command-line utilities, FFmpeg support, and Docker containers 1.

The project has undergone several significant updates throughout its development. In June 2026, Netflix released a new v1 model collection for VMAF 1. Prior milestones include the December 2023 release of libvmaf v3.0.0 and the December 2020 launch of libvmaf v2.0.0, which doubled performance compared to earlier versions 1. Additionally, Netflix open-sourced CAMBI, a tool for detecting artifacts and stripes in video content, in October 2021 1.

The VMAF toolkit supports multiple video quality metrics beyond its primary algorithm, including PSNR, SSIM, and MS-SSIM 1. The implementation is designed for broad accessibility, featuring Windows build support and Docker image availability alongside traditional FFmpeg integration 1.