Technical Details and Advantages The core of this patent lies in utilizing ML models (such as multiple submodules including random forest models) to predict and generate quantization parameters (QP).