Today I officially started my first computer vision project — a food image classifier.
Sounds simple, right? Load some pictures, train a model, done.
Yeah, no. I spent half the day just fighting with TensorFlow instead 😂

🛠️ Environment Setup (aka pain)

I thought I’d just “quickly test some code” but got hit with a combo of errors:

• DLL load failed (classic Windows + TensorFlow issue)
• PIL.Image not found
• scipy not installed
• Oh, and “Please restart your kernel”… about 17 times 🙃

It honestly felt like I was debugging my laptop more than writing code.
But hey — I didn’t quit.

✅ What I Actually Got Done

• Set up a fresh conda environment and linked it to Jupyter
• Used ImageDataGenerator to load and preprocess food images
• Built a ResNet50-based model with transfer learning (no fine-tuning yet)
• Got flow_from_directory() working with 36 classes
• Didn’t throw my laptop out the window (barely)

📅 What’s Next?

• Actually train the model with model.fit()
• Plot accuracy/loss curves
• Save the model (model.h5)
• Try out predictions — maybe I’ll feed it my lunch and see what it thinks?

🧠 Reflection

No big results today — but I did get the pipeline running.
And that’s a win in my book.

Sometimes progress just looks like installing packages, restarting kernels, and not giving up.
Tomorrow, the model gets to do the learning. Today? That was all me.