Advancing ICON with Object Avoidance

Anshul B -

Hello and welcome back to my blog! My name is Anshul Baddi, and I’m excited to share the latest updates on my research project, ICON. Over the past week, I have been working with Professor Ryan Woodward and his fellow students at GCU to learn more about the technology that autonomous cars use. I have been learning about CNN’s from his classes and online courses. Our first phase is object avoidance.

Object Avoidance: The First Phase of ICON

ICON’s initial goal is to classify and identify objects and then decide whether to steer away or stop. This phase ensures the car can react to obstacles using real-time image classification.

Required Hardware

To achieve this, we will be integrating:

  • Raspberry Pi 5 with 8GB RAM: Ensures smooth processing at around 15-30 FPS.
  • Camera: Captures images for classification, and we might use the Rasberry Pi AI camera, which has the MobileNet ML model built into it. 
  • Dual Motor Driver Carrier: Enables the Raspberry Pi to send commands to motors.

Learning Phase: Understanding CNNs

This past week, I have been focusing on Convolutional Neural Networks (CNNs) to understand their role in object detection. A CNN is a type of neural network that consists of layers to classify images. 

Key CNN Components

  • Convolutional Layer: Extracts image features.
  • ReLU Activation: Introduces non-linearity to the images to exemplify features.
  • Pooling Layer (Max Pooling): Reduces dimensions while retaining important features and also removes around 75% of the data 
  • Fully Connected Layer: Maps features to classification labels.

What’s Next?

Now that I have a foundational understanding of CNNs, the next steps involve implementing a CNN-based object detection model into the Rasberry Pi. We will also begin hardware integration once our components arrive. Stay tuned for more updates!

More Posts

Comments:

All viewpoints are welcome but profane, threatening, disrespectful, or harassing comments will not be tolerated and are subject to moderation up to, and including, full deletion.

    ishan_b
    Hi Anshul, this is a fascinating project! I was just wondering in what environment are you planning to test the hardware, indoors, outdoors, etc?
    William T Pieh
    This is going to be an exciting project to work on with you! It's a pleasure having you in our group, and we appreciate your technical experience!
    shreyash_p
    Great thoughts all around, Anshul. I was wondering how you plan to handle real-time processing on the Raspberry Pi - are you planning to optimize your CNN model or use any specific techniques to maintain a stable framerate during object detection?
    Anonymous
    Hey Ishan, that is a great question. Our first field test will be conducted in a classroom environment with chairs, desks, ect.
    Anshul Baddi
    Hey Shreyash, thank you for the insightful question. The most straightforward answer is that we will be optimizing our CNN during the training/testing phase. The CNN optimizes itself by changing the weights and biases between the layers of neurons using forward and backward propagation mathematical functions.
    Ridhi
    What a great project, Anshul! Your insight is fascinating, and I am excited to see your work in the upcoming weeks :)

Leave a Reply

Your email address will not be published. Required fields are marked *