If you have any problems related to the accessibility of any content (or if you want to request that a specific publication be accessible), please contact (firstname.lastname@example.org). We will work to respond to each request in as timely a manner as possible.
Deep Convolutional Neural Networks for MultilabelPrediction Using RGBD Data
Computer Science and Engineering
StatisticsView Usage Statistics
Robotics relies heavily on the system's ability to perceive the world around the robot accurately and quickly. In a narrow setting as in manufacturing this goal is relatively simple. To make robotics feasible in more dynamic settings we must handle more objects, more attributes, and events that may be out of the scope of what a system has been exposed to previously. To this end, the present work focuses on automatic feature formation from RGB-D data, using deep convolutional neural networks, in order to recognize, not only objects but also attributes which are more applicable across objects, including those objects which have not been seen previously. Progress is shown in relation to more standard systems and near real-time classification of multiple targets is achieved.