Machine Learning (Object Detection)

Members:

Evan Law, M.S.

Glen Barcelo, B.S.

Mace Kanda-Matsumoto, B.S.

Introduction

Machine Learning is the new hype of research in Computer Science and a continuing growing trend for research and applications in multiple industries. This project is focused on identifying images (Step 1) such as images of Hawaiian plants, classifying objects (Step 2) such as a bicycle, car or person and combining Steps 1 and 2 for a live analysis using either image or video as the input source. OpenCV was experimented and tested and compared to Tensorflow, much more work and time is needed to implement Steps 1 and 2 for OpenCV due to the open source nature of the project. There are multiple steps that depend on each other which makes it difficult to create a classifier since training a classifier takes time. TensorFlow was created by Google and has many classification and detection algorithms implemented already and is proven to be faster and efficient. There is a lot of support for TensorFlow and already has CNN support which is one of the goals of this project to utilize.

Goals

• Create and train a image classifier based on Hawaiian plants (completed; We have identified that due to Hawaii having very unique plants in the world that it's difficult to encompass the entire list and working with a image data set will be very limited to what we can obtain from Image-Net.)

  • Classify reefs

  • Classify Hawaiian insects

• Create object detection and classification based on general objects such as trees, cars or people (in progress; Exploring into this step of the project, we have identified several algorithms to implement and found some difficulties in following the tutorials to get a working sample. We are continuing all efforts into this area of Machine Learning)

• Able to detect and recognize objects from a source (image, live video or recorded video) (in progress; Once Step 2 of the project is completed we can go to the final phase of the project for completion)

Project Pages

1. Python Tutorials

2. Setup OpenCV and Python on Windows

3. Creating your workspace in Windows

4. Python Files to download and use

5. Creating Samples and Training Cascade to create XML file

6. TensorFlow YOLO v2 Object Detection

Files and Report

You may find all other files not relating to Python or OpenCV at this page.

Documentation for project can be found at this page.

Reports for this project can be found at this page.