Friday, September 2, 2022
10:20 – 11:10 a.m. (CST)
ETB 1020 – **In-person** (or by Zoom for those receiving emails)
Dr. Joshua Peeples
ACES Faculty Fellow & Visiting Assistant Professor, Texas A&M University, Electrical & Computer Engineering
Title: “Statistical Texture Feature Learning for Image Analysis”
Talking Points:
- Convolutional neural networks are biased towards structural textures
- Histogram layer(s) provide statistical context within deep learning models to improve performance
Abstract
Feature engineering often plays a vital role in the fields of computer vision and machine learning. A few common examples of engineered features include histogram of oriented gradients (HOG), local binary patterns (LBP), and edge histogram descriptors (EHD). Features such as pixel gradient directions and magnitudes for HOG, encoded pixel differences for LBP, and edge orientations for EHD are aggregated through histograms to extract texture information. However, the process of designing handcrafted features can be difficult and time consuming. Artificial neural networks (ANNs) such as convolutional neural networks (CNNs) have performed well in various applications such as facial recognition, semantic segmentation, object detection, and image classification through automated feature learning.
A new histogram layer is proposed to learn features and maximize the performance of ANNs for statistical texture analysis. Current approaches using ANNs or handcrafted features do not perform well for some texture applications due to inherent problems within texture datasets (e.g., high intrinsic dimensionality, large intra-class variations) and limitations in methods that use handcrafted and/or deep learning features. The proposed approach is a novel method to synthesize both neural and traditional features into a single pipeline. The histogram layer can estimate bin centers and widths through the backpropagation of errors to aggregate the features from the data while also maintaining spatial information. The improved performance of each network with the addition of histogram layer(s) demonstrates the potential for the use of this new element within ANNs.
Biography
Dr. Joshua Peeples is an ACES Faculty Fellow and Visiting Assistant Professor in the Department of Electrical and Computer Engineering at Texas A&M University. Dr. Peeples received his Bachelor of Science degree in electrical engineering with a minor in mathematics from the University of Alabama at Birmingham. He earned his Ph.D. in the Department of Electrical and Computer Engineering at the University of Florida with Dr. Alina Zare. During his Ph.D. studies, Dr. Peeples developed and refined novel deep learning methods for texture characterization, segmentation, and classification. Dr. Peeples’ current research seeks to extend his dissertation work and explore new aspects such as developing algorithms for explainable AI and various real-world applications in other domains (e.g., biomedical, agriculture). These methods can then be applied toward automated image understanding, object detection, and classification. Dr. Peeples has been recognized with several awards, including the Florida Education Fund’s McKnight Doctoral Fellowship and National Science Foundation Graduate Research Fellowship. In addition to research and teaching, Dr. Peeples is dedicated to service and advocacy for students at the university and in the community.
More information on Dr. Peeples at https://engineering.tamu.edu/electrical/profiles/peeples-joshua.html
Please join on Friday, 9/2/22 at 10:20 a.m. in ETB 1020.