Comparative Analysis of VCM and AhG8 for Machine Vision Applications

Abstract:

This paper presents a performance review of standardization efforts for optimized video compression techniques targeting machine vision. As these efforts are expected to share similar use cases, there is interest in a formal comparison. Considered activities include MPEG's ongoing standard Video Coding for Machines (VCM) and JVET's contributions in AhG8. Multiple simulations across bitrates, encoding configurations, datasets, and machine tasks are performed for fair evaluation. Empirical results show VCM to outperform noticeably in most cases, and significantly in select cases.

Authors:

Perera, A., Adzic, V., Kalva, H., & Furht, B.

Conference / Journal

SSRN 2025