Conference & Workshop Posters

• End-to-end Convolutional Selective Autoencoder for Soybean Cyst Nematode Eggs Detection - Phenotypic Prediction: Image Acquisition and Analysis Workshop, 2016
• Hierarchical Feature Extraction for Efficient Design of Microfluidic Flow Patterns - NIPS Feature Extraction Workshop, 2015
• Early Detection of Combustion Instability by Neural-Symbolic Analysis of Hi-speed Video - NIPS Cognitive Computing Workshop, 2015
• A Knowledge Representation and Information Fusion Framework for Decision Making in Complex Cyber-physical Systems - NSF Cyber-Physical Systems Principal Investigators' Meeting 2015
• Deep Learning for Fluid Sculpting in Microfluidic Platforms - Nvidia GPU Tech Conference 2015
• Generalized Gossip Algorithms for Solving Distributed Optimization Problems - Midwest Controls and Game Theory Workshop 2015
• Pattern Discovery from Large-scale Computational Fluid Dynamic Data using Deep Learning - NSF Data Science Workshop, 2015
• Topology Control in Mobile Sensor Networks using Information Space Feedback - Midwest Controls and Game Theory Workshop 2015
• Understanding Wind Turbine Interactions Using Spatiotemporal Pattern Network - TCIPG Summer School 2015

Videos

LLNet - A Deep Autoencoder Approach to Natural Low-light Image Enhancement: Performance comparison of natural low-light image enhancement using different methods. Note that our framework aims to eliminate noise as well as enhancing contrast simultaneously.
Link to LLNet Code Repository

Decision Making in Cyber-physical Systems: Evolution of belief map for brute force search, convolutional neural network, and augmented Markov decision processes

Early Detection of Combustion Instability with Deep Learning: Combustion instability deteriorates engines and poses safety issues. During unstable combustion, coherent features are present in flame images. We use a selective convolutional autoencoder to learn these unstable features to achieve early detection of combustion instability.

Deep Learning for Fluid Sculpting (Part 1): Demonstration of a CAD tool prototype for predicting pillar sequences (that distort the fluid cross-sectional shape within the fluid channel) based on a hand-drawn flow shape desired by the user.

Deep Learning for Fluid Sculpting (Part 2): Similar to motivation from Part 1, we implemented a Convolutional Netural Network to learn the transformation from one flow shape to another. By learning these transformations, we can incrementally enumerate the pillar sequence.

Deep 3D Convolutional Neural Network Based Design for Manufacturability Framework: Developing a machine learning and GPU acceleration based web tool for Manufacturing Processes.

A day in the Lab

Lab Members @ May 2015: Adrian, Dotun, Collin, Zhanhong, Adam, Jeff, Dr. Sarkar & Kin

Discussion: Although involved in different projects, each lab member continues to contribute ideas to help further research progress.

Zhanhong and Dotun's Birthday @ October 2015. We have new members! Zhanhong, Dotun, Venki, Kin, Dr. Sarkar, Sam & Chao