TensorFlow Object Detection - Object Recognition

This project implements a Object recognition system using TensorFlow and OpenCV. The unsupervised machine learning model accurately identifies and classifies objects in live video streams.

Overview

• Objective:

  • Implement Object recognition using TensorFlow and OpenCV.

• Structure:

  • Utilizes TensorFlow Object Detection API.
  • Trained unsupervised ML model.
  • Live video processing with OpenCV.

Development

• Training Process:

  • TensorFlow and OpenCV used for unsupervised model training.
  • Object Detection API employed for accurate identification and classification.

• Setup:

  • Clone the TFODCourse repository.
  • Set up a virtual environment and install dependencies.

• Data Collection:

  • Use 1. Image Collection.ipynb for image collection.
  • Manual division of images into train and test folders.

• Training and Evaluation:

  • Open 2. Training and Detection.ipynb for training.
  • Tensorboard optionally used for model evaluation.

• Credits:

  • Inspired by the Tensorflow Object Detection Walkthrough by Nick Noach.
  • YouTube channel for a detailed course on custom object detection.

How to Use

1. Compile:

   • Compile the provided code.

2. Setup:

   • Set up/activate the tfod3 virtual environment

3. Running the Model:

   • Open 2. Training and Detection.ipynb.
   • Import dependencies and set up Path variables (Step 1)
   • Load Trained Model From Checkpoint (Step 8)
   • Run Step 9 with modified IMAGE_PATH for static image object detection
   • Run cell 2 of Step 10 to launch Real-Time Object Detection with your camera

4. Updates for Future:

   • Adding more annotated images of each object to testing and training sets for better classification
   • Incorporating GPU to increase speed of training, evaluation, and execution
   • Add more objects to the model
   • Configure the model to run with my VR Headset (so that I can feel like Iron Man)

Enjoy your facial recognition experience!