tianhewu / assessor360 Goto Github PK

Tianhe Wu^1,*, Shuwei Shi^1,2*, Haoming Cai³, Mingdeng Cao², Jing Xiao⁴, Yinqiang Zheng², Yujiu Yang¹

¹Shenzhen International Graduate School, Tsinghua University
²The University of Tokyo
³University of Maryland, College Park
⁴Pingan Group

Tsinghua University Intelligent Interaction Group

🚀 🚀 🚀 Updates:

• ✅ Sep. 22, 2023: Our paper is accepted by NeurIPS 2023!
• ✅ June. 7, 2023: We release the Assessor360 source code.

This repository is the official PyTorch implementation of Assessor360: Multi-sequence Network for Blind Omnidirectional Image Quality Assessment. 🔥🔥🔥 We process all 2D VR IQA dataset MOS files, which can be seen in data folder. If you want to train your model, you can just download the dataset and modify the file in models and data folders.

Click to see different distorted level omnidirectional images. You can also download Insta360 Player to have a immersive experience.

Ground Truth	Distortion 1	Distortion 2	Distortion 3	Distortion 4
MOS (GT)	3.45 (1)	2.95 (2)	1.6 (3)	1.1 (4)
Ours (Assessor360)	0.5933 (1)	0.5213 (2)	0.1220 (3)	0.0120 (4)
MOS (GT)	4.85 (1)	3.25 (2)	2.4 (3)	1.3 (4)
Ours (Assessor360)	0.9566 (1)	0.7263 (2)	0.3495 (3)	0.0748 (4)
MOS (GT)	4.35 (1)	3.35 (2)	2.6 (3)	1.3 (4)
Ours (Assessor360)	0.8194 (1)	0.5481 (2)	0.3583 (3)	0.1165 (4)

Blind Omnidirectional Image Quality Assessment (BOIQA) aims to objectively assess the human perceptual quality of omnidirectional images (ODIs) without relying on pristine-quality image information. It is becoming more significant with the increasing advancement of virtual reality (VR) technology. However, the quality assessment of ODIs is severely hampered by the fact that the existing BOIQA pipeline lacks the modeling of the observer's browsing process. To tackle this issue, we propose a novel multi-sequence network for BOIQA called Assessor360, which is derived from the realistic multi-assessor ODI quality assessment procedure. Specifically, we propose a generalized Recursive Probability Sampling (RPS) method for the BOIQA task, combining content and details information to generate multiple pseudo viewport sequences from a given starting point. Additionally, we design a Multi-scale Feature Aggregation (MFA) module with a Distortion-aware Block (DAB) to fuse distorted and semantic features of each viewport. We also devise Temporal Modeling Module (TMM) to learn the viewport transition in the temporal domain. Extensive experimental results demonstrate that Assessor360 outperforms state-of-the-art methods on multiple OIQA datasets.

We test our model performance in CVIQD, OIQA, MVAQD, IQA-ODI four 2D VR IQA datasets.

You can send emails to OIQA and MVAQD dataset authors to obtain datasets.

Click into the website and download the pretrained model checkpoints.

• Modify "dataset_name" to choose which datasets you want to train in config
• Modify training and validation dataset path

model_name and type_name is the file path for saving checkpoint and log file

python train_oiqa.py

• Modify the path of image "image_path"
• Modify the path of checkpoint "model_weight_path"

python inference_one_image.py 

We would like to thank Xiangjie Sui for his great and patient guidance on this work, Xinzhe Ni for designing the model and DiffBIR published README.md file.

[CVPRW 2022] MANIQA: Multi-dimension Attention Network for No-Reference Image Quality Assessment. (1th place in NR track)

@article{wu2023assessor360,
  title={Assessor360: Multi-sequence Network for Blind Omnidirectional Image Quality Assessment},
  author={Wu, Tianhe and Shi, Shuwei and Cai, Haoming and Cao, Mingdeng and Xiao, Jing and Zheng, Yinqiang and Yang, Yujiu},
  journal={arXiv preprint arXiv:2305.10983},
  year={2023}
}