Street-level Imagery Dataset for the Detection of Informal Vendors in Urban Environment

Informal street vending is a significant aspect of the informal economy and is vital for understanding urban environments. While extensively studied in disciplines like anthropology, economics, and sociology, the global number of street vending remains unclear. Understanding this phenomenon is crucial for analyzing social and economic indicators. However, traditional methods for studying street vending, such as interviews and observational techniques, are often costly and labor-intensive, limiting their scalability and frequency.

To address these challenges, we present the Steet-level Imagery Dataset for detecting Informal Vendors in Urban Environment. This dataset was created using video footage captured by a GoPro action camera mounted on a motorcycle handlebar, producing side-looking images at two-second intervals. The final dataset consists of 2,794 annotated images. To comply with GDPR privacy guidelines, pedestrian faces, and vehicle license plates were anonymized using an open-source Python pipeline powered by the YOLO object detection algorithm.

Annotations were created using the LabelImg tool, where street vendors were identified and labeled with bounding boxes. Each annotation specifies one of three vendor types: "fixed-stall-vendor," "semi-fixed-vendor," and "itinerant-vendor," enabling granular analysis. All annotations are stored in YOLO format for seamless integration into machine learning workflows.

To improve model generalization and address class imbalance, data augmentation techniques were applied. These include geometric transformations such as rotation, flipping, scaling, and shearing, along with spectral adjustments like brightness, contrast, hue changes, blur, and CLAHE (Contrast Limited Adaptive Histogram Equalization).

This dataset is valuable for researchers aiming to develop machine-learning models to detect and analyze informal economic activities. By facilitating scalable and efficient analysis, it supports advancements in urban studies and related fields.