This deep learning project focuses on classifying images of playing cards, utilizing the PyTorch framework. The chosen model architecture for this project is AlexNet (built from scratch, not using transfer learning).
Training Details :
-
Epochs : 50
-
Batch Size : 8
-
Learning Rate : 0.0005, optimized using the Adam optimizer with a weight decay of
1e-4.
Data Preprocessing :
- Images were resized to 227x227 pixels as the sole transformation; no data augmentation techniques (e.g., flipping or rotation) were used.
Validation Methods :
- Validation was performed with Hold-Out Validation Method at the end of each epoch using the validation dataset to calculate loss and accuracy.
- It employs a simple train-validation split where the training and validation sets are predefined and separate. The training loop iterates through the
train_loader, and after each epoch, the model is evaluated on thevalid_loaderto measure validation loss and accuracy.
Training results:
Testing results:
Test Accuracy: 0.8830 Classification Report:
| precision | recall | f1-score | support | |
|---|---|---|---|---|
| accuracy | 0.88 | 265 | ||
| macro avg | 0.91 | 0.88 | 0.88 | 265 |
| weighted avg | 0.91 | 0.88 | 0.88 | 265 |
Confusion Matrix
Custom Image Test Result:
