This repository contains pretrained models and network structure for CU partition at intra-mode.
We use DIV2K dataset. s. To obtain the dataset including intra CU partition modes, all images are encoded by the VVC Test Model (VTM) 7.0 at four Quantization Parameters (QPs) {22, 27, 32, 37}, and the encoder configuration is set to All-Intra mode for getting intra partition samples.
Take an example with 32x32 CU as input.
python main.py --train_path [PATH_TO_DATA] --num_classes 6