This is example use of:
- TensorFlow / Keras
- Convolutional Neural Networks
- Transfer Learning (building on top of existing model, EfficientNetV2B0)
...to reliably tell apart laptops from PC-s in a real photographs.
See it in action at: https://ml.qertoip.com/laptoppc/
Tested on Python 3.10.12.
After cloning the repository, please run:
cd laptoppc
bin/reset-venv # creates .venv and installs dependencies
The trained model is embedded in the project, so you can go straight to running a webapp.
To run webserver locally:
bin/dev-server
Then go to: http://127.0.0.1:5000/laptoppc/
The trained model is embedded in the project, so you can go straight to running these commands.
Always start by activating venv: source .venv/bin/activate
To run some example predictions, run python laptoppc/main_example.py
To visualize NN, run python laptoppc/main_visualize_model.py
To train model from data, run python laptoppc/main_train.py
To recreate data directory, uncomment and run python laptoppc/main_create_dataset.py (not recommended because you will need to manually clean up the dataset).
See model/base_model.png vs model/finetuned_model.png. The final 3 layers differ between base and finetuned model, as expected.