Unilateral facial palsy, a common type of facial paralysis, profoundly impacts individuals' daily functionality and quality of life. The current clinical diagnosis of facial paralysis primarily relies on the subjective judgment of doctors, and the development of automated detection methods faces challenges such as the lack of publicly available facial paralysis datasets and the inability to analyze different facial nerve branches. To address these problems, we propose a new benchmark named UPFG-SG for Unilateral Peripheral Facial Paralysis Severity Grading. First, we establish a dataset with an improved subjective evaluating system to assess the palsy severity of different peripheral facial nerve branches. Second, we propose a new method trained on this dataset which integrates different facial features to rate the facial palsy severity of each facial nerve region. (Our codes will be publicly available soon.) The general technical approach is illustrated in the following figure. The study was approved by the Ethics Committee of Peking Union Medical College Hospital (Reference No. K22C0265, Approval No. I-22PJ052).

Overview of the UPFP-SG dataset construction and the UPFP-SG method for facial palsy severity grading.
(Note: The facial images shown are synthesized images for illustration purposes and do not correspond to any real individual.)

Under the guidance and assistance of medical professionals in Peking Union Medical College Hospital, we established the first UPFP multi-level evaluation dataset targeting different facial nerves, which is identified as the UPFP-SG dataset. The construction of this dataset in the study of this article was approved by the Ethics Committee of Peking Union Medical College Hospital under reference number K22C0265. We collected and organized facial paralysis video image data according to the established collection specifications and grading assessment criteria. Facial nerve functional videos were collected from 59 patients with facial nerve disorders, representing a diverse group in terms of age, gender and condition severity. The dataset was annotated by several doctors in a blind manner according to the improved facial nerve peripheral branch grading system proposed in this paper.

We make UPFP-SG dataset available for academic research purposes. However, as it involves a lot of personal privacy, we only send the data to approved researchers. To access the dataset for research (noncommercial use), please send an email to drtzhang@126.com. Please note that any commercial use of the dataset is prohibited. We are currently curating and standardizing the UPFP-SG dataset, and plan to release the complete version publicly by December 2025.