RibCageImp: A 3D Deep Learning Framework for Rib Cage Implant Generation

Prepared Data from scratch using RibFrac dataset. I have artificial created defected regions into the Ribcage and generated corresponding implants as labels. In first phase, developed a deep learning model that takes CT scans as input and generates 3D implants for rib cage generation.

GitHub | Project Page

Image Generation Using Conditional GANs

Generated dermoscopic lesion images using Conditional Generative Adversarial Networks (cGANs). Implemented three different architectures for cGANs, including those with paired sketches and actual images, as well as with sketches alone. Obtained a FID score of 0.22 with the second architecture.

GitHub

Image Segmentation Using K-Means Clustering and the Ratio Cut Algorithm

Developed K-Means and Ratio Cut algorithms from scratch for image segmentation, comparing their performance with the Silhouette Score. For a 64x64 image with 3 clusters, K-Means achieved a score of 0.53 (indicating optimized results) versus 0.25 for Ratio Cut.

GitHub

Corners Detection in Images

Implemented the Harris Corner Detection algorithm from scratch and compared its performance with the OpenCV library’s implementation. Achieved approximately 80-90% qualitative accuracy in both cases.

GitHub