A multi-view annotation tool for people detection evaluation

摘要: In this paper we introduce a novel multi-view annotation tool for generating 3D ground truth data of the real location people in scene. The proposed allows user to accurately select occupancy by aligning an oriented rectangle on plane. addition, height can also be adjusted. order achieve precise is aided video frames multiple synchronized and calibrated cameras. Finally, easily converted 2D image positions using available calibration matrices. One key advantage technique that different methods compared against each other, whether they estimate world position or camera images. Therefore, defined two error metrics, which quantitatively evaluate estimated positions. We used annotate publicly datasets, evaluated metrics state art algorithms.