Below are my end-to-end projects demonstrating my skills in detection, segmentation, tracking, generation, and 3D morphing. Each project has a detailed PDF notebook attached for full explanation and results.
In this project, I used YOLO to detect helmets from real-world images. I can build custom object detectors for any category.
📄 See full work: Project1.pdf
Tracked vehicles using YOLO for detection and ByteTrack for tracking and counting. Used for traffic analytics and real-time monitoring.
📄 See full work: Project2.pdf
Performed semantic segmentation to extract dogs from images. Also trained a SegFormer model on my custom dataset. 📄 See full work: Project3a.pdf 📄 See full work: Project3b.pdf
Applied instance segmentation techniques for object-level masking and visualization. Explored try-on systems for segmented objects.
📄 See full work: Project4.pdf
Created an image captioning model that generates natural text descriptions from images. Can be customized for your own dataset.
📄 See full work: Project5.pdf
Used diffusion models to generate and enhance images from text and image inputs. Demonstrated prompt-based product generation and image upscaling.
📄 See full work: Project6.pdf
Performed depth map estimation from single images using pretrained depth models. Applied color grading filters similar to editing apps.
📄 See full work: Project7.pdf
Developed a model to remove backgrounds from images containing people and objects. Ideal for e-commerce, portraits, and photo editing tools.
📄 See full work: Project8.pdf
Built a system that morphs a neutral 3D face mesh based on the expressions detected in a user’s 2D image (e.g., smiling).
📄 See full work: Project9.pdf
Attempted to generate a 3D avatar by combining the user’s 3D head shape (from 2D input) with a pre-made body template.
📄 See full work: Project10.pdf
💡 All projects demonstrate real computer vision and 3D AI capabilities — from classic detection to advanced 3D morphing — fully implemented, tested, and documented.