Week 5: Progress on ICON
Anshul B -
Hello and welcome back to my blog! My name is Anshul, and this week was relatively quiet as all students were on spring break, so no physical meetings took place at my site placement. However, we had an online meeting on Discord to discuss our progress in the machine learning course. We also planned to get a battery for the Raspberry Pi and finally acquired a webcam for the CNN training.
Independent Research
This week, I focused on learning more about Depth from Defocus (DFD) and how we will implement it. Additionally, I worked on debugging my CNN, facing issues like inaccurate object predictions and overfitting of my model. In this blog post, I mainly want to focus on our code’s key components.
Key Components of Our Code
- Train the CNN Model
- Prepare and label the dataset.
- Train and evaluate the model, addressing overfitting.
- Integrate Depth from Defocus (DFD) for Distance Measurement
- Capture defocused images and process blur differences.
- Estimate depth maps.
- Develop the Decision-Making Algorithm
- Use CNN and DFD data to determine obstacle distances.
- Define movement rules (turn, slow, stop).
- Implement Steering Commands on Raspberry Pi
- Connect the hardware and write control code.
- Test in a classroom environment.
That’s all for this week! Stay tuned for more updates as we move forward with implementing these components. Also, I have attached a picture of our current setup of ICON down below!
