Task
The goal was to train a model that assigns an image to one of the ten CIFAR-10 classes and to push accuracy well above the baseline.
Solution
I built a convolutional network in Keras and TensorFlow, prepared and augmented the data, added convolutional blocks with regularization against overfitting and tuned the hyperparameters. Training and metrics ran in Jupyter, and the final model is saved for inference.
What was done
- Built a convolutional network in Keras and TensorFlow for 10 classes
- Raised accuracy above the baseline through hyperparameter tuning
- Added convolutional blocks with regularization against overfitting
- Augmented the data and saved the trained model for inference
Result
A working model with stable test accuracy and saved weights. It can be plugged in for image recognition without retraining.
Technologies
PythonKerasJupyter
