小米发了套新评估指标RC-S和RC-T，专门测视频物体移除效果，比PSNR那些老指标靠谱，搞视频编辑的可以看看。
小米MiLM Plus团队发布PROVE，一套面向视频物体移除的感知对齐评估指标，包含RC-S和RC-T两个新指标。现有PSNR、SSIM、LPIPS等指标在扩散模型输出上常给出错误排序，因为物体移除是病态的一对多任务。PROVE引入真实世界视频基准，通过感知对齐方式评估移除质量。该工作旨在解决评估指标滞后于模型能力的问题。
Xiaomi’s MiLM Plus Releases PROVE: Perception-Aligned Object Removal Metrics RC-S and RC-T With a Real-World Video Benchmark
Object removal models have improved faster than the metrics used to judge them. Diffusion erasers now reconstruct shadows, reflections and occluded structure convincingly, yet PSNR, SSIM, LPIPS, ReMOVE and CFD frequently rank their outputs the wrong way. The root cause is structural: erasure is an ill-posed, one-to-many task, so no single ground truth exists to […] The post Xiaomi’s MiLM Plus Releases PROVE: Perception-Aligned Object Removal Metrics RC-S and RC-T With a Real-World Video Benchmark appeared first on MarkTechPost .