Haider Ali, JHU Center for Imaging Science – “Feature Selection and Fusion for 3D Object Category Recognition”

When:
November 16, 2017 @ 10:30 am – 11:45 am
2017-11-16T10:30:00-05:00
2017-11-16T11:45:00-05:00

Location

Hackerman B-17

Abstract

Deep learning methods have received lots of attention in research on 3D object recognition. Due to the lack of training data, many researchers use pre-trained convolutional neural networks (CNNs) and either extract the output of one of the last layers as features or fine-tune the networks on their data. Due to the extraordinary features that can be obtained from a CNN, we intend to use them for 3D object recognition in the task of robotics using active learning and the Mondrian forest classifier. We achieve superior results with a method that fine-tunes a CNN before feature extraction for RGB data. Combined with extracted features from depth data and reducing the features’ dimensionalities, we improve the state-of-the-art accuracy on the University of Washington RGB-D Object dataset in the standards offline case, using a support vector machine (SVM). Instead of SVM as a classifier, we use active learning and the Mondrian forest, an online classifier, which can be updated over time once more data is available. Additionally, in our earlier work we present a novel combination of depth and color features to recognize different object categories in isolation. We also investigate the effect of domain change by training on RGB-D Object dataset and testing on DLR-RGB-D dataset. In our experiments we show that a domain change can have significant impact on the model’s accuracy, and present results for improving the results by increasing the variability of the objects in the training domain.

Bio

Dr.Haider Ali is currently serving as an Associate Research Scientist at The Center for Imaging Science (CIS), Johns Hopkins University.
Before joining CIS he worked as a Senior Researcher at the Institute of Robotics and Mechatronics (RM), German Aerospace Center (DLR). His research is primarily focused on developing efficient 3D object and activity recognition methods for real time robotic applications. He received his Bachelor of Science in Computer Science from Bahauddin Zakariya University one of Pakistan’s major universities in 1998. After that he served several multinational IT companies in Pakistan as Software Engineer and Project Consultant until 2004. Thereafter he planned to pursue a master degree in software technology from Leuphana University of Lueneburg and graduated in 2006. He received his Ph.D. from Technical University of Vienna in 2010.

Host

Greg Hager

Video

Watch seminar video.

Back to top