Object Detection using Transfer Learning with R-CNN model

In this project I am tasked with identifying an object (in this case, a cellphone) in a given scene and compute the bouding box along with the center pixel coordinates as an output

Use the following instructions to build the code

Dependencies

• Ubuntu 18.0x - Operating System
• Python 3.7.1 - Programming Language Used
• Tensorflow 1.15.3 - Deep Learning and R-CNN Model
• Keras 2.0.9 - Interface for Tensorflow library
• h5py 2.10.0- Model Saving and Checkpoint Creation
• Mask R-CNN model - Deep Learning Model for transfer learning

Please download the R-CNN model weights for COCO dataset into the current working directory from the following link: https://github.com/matterport/Mask_RCNN/releases/download/v2.0/mask_rcnn_coco.h5

Inside the Repo

The repository contains python scripts to train and test the model. It also contains the dataset on which the model was trained.

Files

• find_phone.py : Python script to test the model.
• train_phone_finder.py : Python scrip to train the model.
• mask_rcnn_coco.h5 : Weights of the network trained for COCO dataset.
• readme.md

Folders

• Mask_RCNN: Deep Learning Model used for Learning. Please Install the R-CNN model from the matterport/Mask_RCNN github repo
• dataset: Auto-generated folder with images and annotations
• find_phone: Dataset folder

Training Script

• train_phone_finder.py : takes a single command line argument which is a path to the folder with labeled images and labels.txt

python3 train_phone_finder.py ~/find_phone

Testing Script

• find_phone.py : takes a single command line argument which is a path to the jpeg image to be tested.

 python3 find_phone.py ~/60.jpg

Output

Make sure the weights created by the training script is located in the above mentioned directory of the test image. Below is the expected output of the testing script.

Phone in image 60.jpg is located at x-y coordinates given below.

0.1449 0.3558

A comparison of the predicted images and the actual images for object detection:

The centroid of the object along with the bounding box can be seen as: