Add finetuning script and initial implementation

[Devasy]

Add initial implementation of DeepTurne package for fine-tuning CV-based deep models using Siamese architecture with triplet loss function.

• README.md

  • Add project description, installation instructions, and usage examples.

• deepturne/init.py

  • Initialize the deepturne package.

• deepturne/triplet_loss.py

  • Implement custom triplet loss function.

• deepturne/backbones.py

  • Add support for model backbones including ResNet, VGG, Inception, MobileNet, EfficientNet, Facenet, and Arcface.
  • Load pre-trained weights for Facenet and Arcface models.

• deepturne/data_preprocessing.py

  • Implement data preprocessing steps including normalization, resizing, augmentation, and triplet creation.

• deepturne/evaluation_metrics.py

  • Implement evaluation metrics including accuracy, precision, recall, F1 score, mean average precision (mAP), ROC-AUC, and triplet loss.

• deepturne/datasets.py

  • Add support for downloading and loading basic datasets for fine-tuning, such as LFW, using TensorFlow Datasets (TFDS).

• deepturne/test_finetuning.py

  • Implement a script to test the fine-tuning of at least one model for a few epochs.

---

For more details, open the Copilot Workspace session.