June 4, 2025
Day 7 – Diving into the codebase
What I Learned
Today, Mr. Pelumi gave us the project demo codebase. Since we’d be using TensorFlow for the project, I spent a good amount of time refreshing my limited TensorFlow skills. My colleagues faced some issues running the model on their machines, so I helped them. I also provided them with some books and study materials from my previous semester, which could help them understand the basics of ML.
The pre-trained model used in the codebase was a simple MobileNetV2 model, so I read a short article on it to understand its architecture. I also studied transfer learning and how to implement it. Then, I tuned some hyperparameters like the learning rate and optimizer, and implemented early stopping and a learning rate scheduler to see if they’d improve the accuracy. They did, although only slightly. Additionally, I reviewed my past deep learning projects to identify components that could be incorporated into our current project. After I get back home, I plan to continue studying transfer learning.
Blockers
No issues faced.
Reflection
I’d say today was a productive day as I not only refreshed my TensorFlow skills but also helped my colleagues. Helping them not only fixed their issues but also reinforced my own knowledge and enhanced my image as a supportive person, which will encourage more cooperation and coordination in the coming days. Regarding hyperparameter tuning, although I didn’t make a significant improvement, it was still a learning experience as I explored more optimizers and their parameters, and I’d call that a win. Overall, I feel more confident in my TensorFlow skills and am excited to apply what I’ve learned to our project.