Spaces:

---

xinyiW915

/

DIVA-VQA

like 1

Sleeping

Project Name: DIVA-VQA: Detecting Inter-Frame Variations in UGC Video Quality

Project Type: Academic Research Project

Project Background:
DIVA-VQA builds upon our previous model, ReLaX-VQA. The ReLaX-VQA project is publicly available on GitHub and received positive attention, including an invitation from the Hugging Face team to build a demo based on our repository:
🔗 GitHub Issue

DIVA-VQA is our new model that enhances video quality assessment by focusing on inter-frame variations. The project will be fully open-sourced and published soon.

Why We Need GPU:

• The current CPU-based demo on Spaces (2 vCPU, 16 GB RAM) is significantly limited in performance, especially with high-resolution or long-duration videos.
• DIVA-VQA relies on deep learning inference across multiple video frames, which is computationally intensive and slow without a GPU.
• With GPU support, we can greatly improve user experience, and enable more extensive testing.

Open Source: Yes — both the code and demo will be fully open-sourced soon.

Demo Link:
🔗 https://huggingface.co/spaces/xinyiW915/DIVA-VQA

We kindly request GPU access through the ZeroGPU grant program to help us. Thank you for your support! 🙏