Date: 30 Oct 2014 (Thursday)
Time: 10.30am - 11.30am
Venue: Meeting room, Mimos Lab, SB Ground Floor, FES, UTAR Kuala Lumpur Campus.
Title: Convolutional neural networks for gender recognition
Most computer vision methods relies heavily on hand-crafted features such as SIFT, LBP and HOG. Recently, deep learning methods, characterized by neural network architectures with many hidden layers, have become popular. In this paradigm, discriminative features are learnt by the network, thus enable it adapt to a different problem domain. The availability of parallel computing power at reasonable cost in the form of consumer graphical processing units has resulted in a resurgence of interest in such neural networks, as large architectures can now be trained on datasets consisting of many thousands of images. In particular, deep convolutional neural networks has achieved state-of-the-art results in tasks such as traffic sign classification, large scale object recognition, pedestrian detection and action recognition. In this project, we study convolutional neural networks for the problem of gender classification.
Speaker: Ng Choon Boon, Lecturer, UTAR.