This dataset was developed to facilitate the training and evaluation of computer vision models for obstacle detection in robotics. Images were captured in a controlled environment, specifically on the competition floor used in the RobotAtFactory 4.0 (RAF) competition, to resemble the expected future application scenarios of the algorithm closely. The data was collected using a mobile platform with two cost-effective cameras: the Raspberry Pi Camera Module v2 (RaspCamV2) and the ESP32-CAM (EspCam). The dataset comprises 2,907 images at a resolution of 640x480 pixels. Among these, 1,638 images contain annotations of obstacles represented by mannequins, while 1,269 images depict backgrounds without them. Annotations were performed using the LabelImg software, ensuring compatibility with the YOLO model. This dataset is well-suited for object detection tasks in robotics, providing a diverse array of scenarios that support the training and benchmarking of machine learning algorithms, particularly in environments similar to those encountered during the RAF competition.
