"Look at this one" detection sharing between modality-independent classifiers for robotic discovery of people

摘要

With the advent of low-cost RGBD sensors, many solutions have been proposed for extraction and fusion of colour and depth information. In this paper, we propose new different fusion approaches of these multimodal sources for people detection. We are especially concerned by a scenario where a robot evolves in a changing environment. We extend the use of the Faster RCNN framework proposed by Girshick et al. [1] to this use case (i), we significantly improve performances on people detection on the InOutDoor RGBD People dataset [2] and the RGBD people dataset [3] (ii), we show these fusion handle efficiently sensor defect like complete lost of a modality (iii). Furthermore we propose a new dataset for people detection in difficult conditions: ONERA.ROOM (iv).