Revisiting Video Quality Assessment from the Perspective of Generalization

Code for CVPR 2025 submission paper "Revisiting Video Quality Assessment from the Perspective of Generalization".

Environment

• Python (3.11.5)
• Pytorch (2.1.2)
• torchvision (0.16.2)
• CUDA (12.0)

Content

• ./fastvqa: Code for FAST-VQA-AWP.
• ./sama: Code for SAMA-AWP.
• ./simplevqa: Code for SimpleVQA-AWP.

Run

• FAST-VQA-wd

cd ./fastvqa
CUDA_VISIBLE_DEVICES='0' python train.py --wd 0.0005

• FAST-VQA-AuA

cd ./fastvqa
CUDA_VISIBLE_DEVICES='0' python train.py --aua

• FAST-VQA-RWP

cd ./fastvqa
CUDA_VISIBLE_DEVICES='0' python train.py --rwp --gamma 0.0001

• FAST-VQA-AWP

cd ./fastvqa
CUDA_VISIBLE_DEVICES='0' python train.py --awp --gamma 0.0001

• SimpleVQA-AWP

cd ./simplevqa
CUDA_VISIBLE_DEVICES='0' python train_LSVQ.py --awp --gamma 0.0001

• SAMA-AWP

cd ./sama
CUDA_VISIBLE_DEVICES='0' python train.py --awp --gamma 0.0001

Reference Code

[1] SimpleVQA: https://github.com/sunwei925/SimpleVQA

[2] FAST-VQA: https://github.com/VQAssessment/FAST-VQA-and-FasterVQA

[3] SAMA: https://github.com/Sissuire/SAMA/tree/main