Simultaneous object detection and localization using convolutional neural networks

摘要

Nowadays deep learning is considered as a trendy technique in the computer vision domain. It becomes a pioneer in its main tasks as object classification, object localization, and object detection. Therefore it gave amazing results and records. In this paper, we propose a new approach to identify and localize objects, simultaneously, in a given scene using Convolutional Neural Networks (CNNs). We propose an end-to-end approach for object detection and pose estimation by formulating bounding boxes containing the targeted object and their pose. Our method is based on two main steps, i) produce Bounding boxes on the training images for generating the pose coordinates of each object in the scene and, ii) detect and localize simultaneously each object present in image during the testing step. The contribution performance is assessed on two datasets, Washington RGB scene dataset and LIMIARF dataset that is constructed in our laboratory. We demonstrate that our proposal is able to obtain high precision and reasonable recall levels.