People and Topics

2019 Fall Schedule

All meetings are held in EE 013.








Software Engineering

Image Database


Human Behavior




Embedded 2


Embedded 1



VIP lecture (EE 129)

Due to the large number of team mebers, it is not possible changing the regular meeting time. If your schedule does not fit into a particular team, you have to move to another team.


Forest Inventory Analysis

Use computer vision to calculate the sizes of trees (called diameter at breath height, or DBH), recognize the species of trees, and their locations. For Fall 2019, the team has two major goals: (1) handle multiple trees in a single frame and (2) handle trees in a nautral forest.

The following images show the result from a distance sensor and the tree image (before and after denoising).

forest03 forest04 forest05

The following images show how the team measures the diameter at breath height.

forest00 forest01 forest02

Readings for new members:

Analysis of Drone Video

This project creates computer vision solutions recognizing objects captured by cameras mounted on drones. In Fall 2019, the team will create a set of video clips for the following purposes:

  • Construct three-dimensional geometries of objects: The video clips will capture cardboard boxes of different sizes, together with a wide range of objects and several with known sizes.

  • Detect and track multiple moving objects: The clips include moving objects. The drone itself is also moving. The purpose is to correctly identify these objects and track their movements.

  • Segmentation: Create pixel-wise labels of different objects.

  • Re-identify people: Determine whether the same person has been before.

This project is supported by NSF CNS-1925713

Readings for new members:

Embedded Vision 1

Recent progress in computer vision has focused primarily in general-purpose object detection using datasets with many (hundreds) categories of objects (such as humans, dogs, vehicles, furniture, buildings, etc.). For many applications, however, the number of possible objects can be limited. For example, inside an airport terminal, elephants or eagles are not expected. This project will use computer graphics to synthesize images and videos of these scenarios. The synthesized data is used to train computer vision running on embedded systems (also called edge devices). Doing so can reduce network traffic and make the system more scalable. Moreover, sensitive information (such as human faces) may be detected and protected before the data leaves the cameras.

Readings for new members:

Analyze Human Behavior in Video

The purpose of this team is to use real-time video analytics to detect dangerous behavior or safety violation in workplace (such as factories), raise alerts to prevent injury, or provide post-event analysis to prevent future occurrences. In Fall 2019, the team will focus on solving these problems in an indoor environment with multiple cameras:

  • Where are the people (including re-identifying the same person in different cameras)?

  • Where does each person face?


Software Engineering for Machine Learning

This project creates a process for developing reproducible software used in machine learning. In Fall 2019, the team’s focus is to create tools that faciliate code review. The tools analyze the histories of version control repositories and automatically identify possible defects within a pull request. The tools will also collect metrics relating to the code review.



Computer vision is still not perfect and humans outperform computers in many situations. This team builds computer tools (human interfaces) for humans to identify unexpected properties (called “bias”) in data used to train computer programs. These tools are computer games and the players (crowds) describe the characteristics in the data.

Reading for new members:

crowdsource03 crowdsource02

crowdsource05 crowdsource04

Image Database

This system integrates computer vision and database. After the objects in images are detected, the information is stored in a database so that it is searchable. The team has built a prototype of the system processing multiple video streams simultaneously. The team will focus on improving the performance (scalability) for lower latency as well as investigating new storage systems.

Reading for new members:

Embedded Vision 2

This project investigates computer vision solutions that can perform the following tasks in an embedded computer (small enough to be inside a typical camera)

  • Obtain aggregate information (such as the number of people and their genders)

  • Detect faces

  • Encrypt the faces before sending the data to storage

The sensitive data (faces) never leaves the camera. Only authorized people with the decryption key can see the faces. The concept is illustrated below.


Readings for new members:


David Michael Barbarash

Landscape Architecture, Purdue

Dave Capperlleri

Mechanical Engineering, Purdue

Shuo-Han Chen

Institute of Information Science, Academia Sinica

Yung-Hsiang Lu

Electrical and Computer Engineering, Purdue

Vinayak Rao

Statistics, Purdue

Guofan Shao

Professor, Forestry and Natural Resources, Purdue

George K. Thiruvathukal

Computer Science, Loyola University Chicago.

Mark Daniel Ward

Statistics, Purdue

Keith E. Woeste

Forestry and Natural Resources, Purdue

Ming Yin

Computer Science, Purdue


Graduate Students

Abhinav Goel: Doctoral Student, Improve Neural Networks’ Energy Efficiency

Sara Aghajanzadeh: Master Student, Detect Faces and Protect Privacy

Ryan Dailey: Master Student, Discover Network Cameras

Undergraduate Students and 2019 Summer Teams

Image Database

Shunqiao Huang


Hojoung Jang

Akshay Pawar

Aditya Chakraborty

Lucas Wiles

Dataset Distinctiveness

Identify the specific features (called distinctiveness) of different visual dataset. Use one dataset with many labels to help train machine models for another datasets with few labels.

Ashley Kim


Damini Rijhwani

Kirthi Shankar Sivamani

Esteban Gorostiaga

Shuhao Xing

Crowdsourcing for Data Bias

Use crowd (i.e., humans) to identify unintentional biases in visual datasets.

Xiao Hu


Haobo Wang


Kaiwen Yu

Anirudh Vegesana

Somesh Dube

Forest Inventory

Use computer vision to calculate the sizes of trees (called diameter at breath height, or DBH).

Nick Eliopoulos

Yezhi Shen

Yuxin Zhang

Vaastav Arora

Minh Nguyen

Human Behavior

Track human activities and understand how they use designed space.

Mohamad Alani

Peter Huang

Dhruv Swarup

Chau Minh Nguyen


Video by Current and Former Members