A dual-view image dataset for suspicious illegal-parking vehicle localization around university roads

This dataset contains 584 de-identified front-view and top-view images from 277 matched vehicle groups collected around university roads in Changqing University Town, Jinan, Shandong, China. Front views were captured with a Huawei nova 12 Pro smartphone. Top views include 3.7 m carbon-fiber selfie-stick timed overhead photographs and selected frames from DJI Flip 2 UAV videos recorded along rows of parked vehicles. The dataset provides single-class bounding-box annotations for vehicle localization in suspicious illegal-parking scenes in YOLO and COCO formats, together with train, validation, and test splits, image metadata, annotation metadata, SHA-256 checksums, privacy-review records, and quality-control reports. Public-release filenames are de-identified, EXIF metadata has been removed, and license plates were manually checked and masked in the public candidate images. The dataset is intended for vehicle localization and road-context-aware parking-scene research. It should not be interpreted as a complete legal/illegal parking classification dataset because systematic legal-parking negative samples and pixel-level road-context labels are not included.