Using Mask R-CNN Modeling for Object Detection and Its Various Applications
By: Nicole Adelson
Mentors: Dr. Edward Currie and Dr. Yimin Zhao
Objectives
● To explore Mask Region-based Convolutional Neural Network (Mask R-CNN) to develop a model that uses pixel-level segmentation to aid in detecting specific objects in an image
● To identify applications that can use the method of object detection in a variety of fields
Main Python Libraries
● Tensorflow
● Keras
Chosen IDE
● PyCharm
Supporting Python Libraries Including, but not limited to:
● NumPy
● Skit-Image
● Matplotlib
Step 1: Establishing the Dataset
Step 2: Parsing the Data
Output:
Dimensions of Each Bounding Box
{
{
Width and Height of the Image
Step 3: Developing the Object
Output:
Step 4: Testing Object - Algorithm
Step 4: Testing Object - Output
In another tab: In console:
Dimensions of the displayed bounding boxes
{
Step 6: Training the Mask R-CNN Model for the Dataset - Preparation
Necessary file for weights:
Step
7: Training the Mask R-CNN Model for the Dataset - Algorithm
Step 7: Training the Mask R-CNN Model for the Dataset - Output
Step
8: Evaluating the Mask-RCNN ModelAlgorithm
Step 8: Evaluating the Mask-RCNN Model - Output
→ 90.6% accuracy
→ 94.3% accuracy
Step 10: Detecting Objects in New PhotosAlgorithm Pt.1
Step 10: Detecting Objects in New Photos -
Algorithm Pt.2
Step 10: Detecting Objects in New PhotosOutput
Applications for the Mask R-CNN Model
● Speech recognition
● Text recognition
● Fraud detection
● Stock market projections
● Medical symptom identification
● And much more!
Source for Model Algorithm Inspiration
Brownlee, Jason. “How to Train an Object Detection Model with Keras.”
MachineLearningMastery.Com, 1 Sept. 2020, machinelearningmastery.com/how-to-train-an-object-detection-model-with-keras/.
Thank you for listening!