Week 4: Advancing ICON and Overcoming Challenges

Anshul B -

Hello and welcome back to my blog! My name is Anshul, and this past week has been an exciting one for our research team, as we made significant progress. From deepening our understanding of machine learning fundamentals to successfully debugging our CNN, we are progressing with the ICON project.

Team Progress This Week

This week, my teammates focused on catching up with the mathematical foundations behind machine learning, specifically linear algebra concepts crucial to understanding how neural networks work. Additionally, one of my teammates successfully accessed our Raspberry Pi 5, meaning we can now begin coding directly into the hardware (I have included a picture of this below). Another thing we achieved was presenting the ICON project at the CAIR (the AI club at GCU) meeting, where we outlined our research goals and explained how ICON will function as a vision-based autonomous navigation system.

Independent Work: Presentation & CNN Debugging

Most of my time this week was spent on debugging my CNN model, which I finally got to work! The main issue was that I needed to install TensorFlow-Metal on my Mac to enable GPU acceleration, allowing me to properly train my CNN. Once I resolved this, my model was able to train efficiently, marking a step forward because we can classify objects now. Additionally, my time was spent creating and delivering the ICON presentation, which covered our objectives, methodologies, and the role of CNN in ICON’s navigation system.

Moving Forward: Expanding Our Work

With our CNN successfully built and access to the Raspberry Pi secured, we can now begin tinkering with the hardware and testing our neural network in a real-world environment. We are just waiting on the camera and battery to arrive. Until then, we can shift our focus to implementing other key components of our system, including the Depth-from-Defocus (DFD) method to estimate distances. Overall, the next steps involve refining our CNN model, integrating it with the Raspberry Pi, and beginning development on depth estimation techniques.

Thank you for following our journey! Stay tuned for next week’s update.

IMG_5044

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.

    camille_bennett
    Hi Anshul, sounds like you are making great progress. What was it like presenting at CAIR? Did they have any questions that surprised you?
    Anshul Baddi
    Our presentation at CAIR was brief due to a time constraint with other groups presenting. However, it was a very in-depth explanation of how we are going to implement ICON and our future goals. Although none of the questions surprised me, there were a lot of questions about how we will integrate our code onto the Rasberry Pi.

Leave a Reply

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