could you introduce some about the dataset size, training environments, training speed

[icegomic]

could you introduce some about the dataset size, training environments, training speed

[tommyhq]

could you introduce some about the dataset size

There are 3422 distorted videos in the training set. These clips were compressed from 176 references, constant QP in range [8, 47]. Clips with extremely small or large QP are discarded for training.

training environments

The code is developed with PyTorch 1.4.0, Python 3.5.6, cuda 9.0

training speed

It takes almost a week to train for 100 epochs with 4 P40 GPU.

[icegomic]

could you introduce some about the dataset size

There are 3422 distorted videos in the training set. These clips were compressed from 176 references, constant QP in range [8, 47]. Clips with extremely small or large QP are discarded for training.

training environments

The code is developed with PyTorch 1.4.0, Python 3.5.6, cuda 9.0

training speed

It takes almost a week to train for 100 epochs with 4 P40 GPU.

Very thanks for your details!