A large-scale, expert-annotated dataset of real clinical periapical radiographs, covering pixel-level annotations of nine anatomical structures and lesions, including teeth and alveolar bone.

NKUT was jointly developed by the NKICS Laboratory and Tianjin Stomatological Hospital. It uses pediatric CBCT data collected and annotated by experts from real clinical settings, providing pixel-level annotations of bilateral wisdom tooth germs, second molars, and parts of alveolar bone.

DentalDS data were collected from patients who visited the Department of Pediatric Orthodontics at Tianjin Stomatological Hospital between January 2021 and May 2024. It includes 2358 carefully annotated pediatric panoramic dental radiographs used for pediatric dental development grading.

The DDR dataset contains 13,673 anonymized fundus images from 9,598 patients across 147 hospitals in 23 provinces in China (2016–2018), with DR grading, pixel-level, and lesion bounding box annotations labeled by expert ophthalmologists.

The OIA-ODIR dataset is derived from a private clinical database with over 1.6 million images from 487 hospitals across 26 provinces. Disease categories are unified for labeling, and annotations are completed through multi-expert consensus and arbitration, ensuring high reliability.

The NKUP dataset is collected in an open environment at Nankai University using 15 cameras. Pedestrian images are cropped from videos and sampled every 25 frames to form an image-based dataset. It includes 5,336 training images of 40 identities, 332 query images, and 4,070 gallery images of 67 identities, with an average of 2.2 outfits per person.

NKUP+ is a long-term public person re-identification dataset with over 40,000 images captured by 29 indoor and outdoor cameras, covering multiple views and appearances. The test set is divided into same-appearance, significant cross-appearance (>3 months), and moderate cross-appearance galleries for comprehensive evaluation.