I developed a deep learning

I developed a deep learning image segmentation system to automatically identify lung regions in chest X-ray images using a UNETR-inspired architecture, combining Vision Transformers with CNN-based decoding.

The model processes X-ray images as patches through a Transformer encoder, using self-attention to learn global image features. A U-Net-style CNN decoder then reconstructs these representations into pixel-level lung segmentation masks using upsampling and skip connections.

I built the complete pipeline including image preprocessing, model architecture, training, validation, evaluation and inference visualization. The model achieved approximately 92% validation accuracy, with segmentation performance evaluated using Intersection over Union (IoU).

This project demonstrates my ability to develop end-to-end computer vision and deep learning solutions, particularly for medical image analysis.

Key skills: Python · Deep Learning · Computer Vision · UNETR · Vision Transformers · CNNs · TensorFlow/Keras · Medical Image Segmentation