People’s detection and tracking methods are susceptible to occlusions and overlapping problems in a multi-camera environment. Such issues are persistent when you mount the camera at a low height. Detection and tracking performance can be improved by combining visual inputs from synchronized camera setups. In this paper, we provide a novel large-scale, high-resolution multi-camera dataset. The dataset has been captured with four stationary cameras with the common region of interest in a university campus where a swarm of spontaneous people are standing and walking. Including frames, we present extrinsic and intrinsic calibration parameters and four sequences of annotated frames for detection at two frames per second. As a result, almost 150 bounding boxes delimit each person present in the region of interest.This dataset can be used for multiple computer vision tasks like pedestrian detection, classification and person reidentification etc.