Computer Vision Projects
Project 1: Image Processing & Color
Built a comprehensive image processing pipeline implementing hybrid image creation, Gaussian and Laplacian pyramids, and multi-resolution blending. Developed algorithms for color space transformations (RGB, YCbCr) and created high-quality hybrid images by manipulating high and low frequency components.
View Project →
Project 2: Fun with Frequencies
Implemented advanced frequency domain operations using Fast Fourier Transform (FFT). Created custom filters for image denoising, sharpening, and blurring. Developed algorithms for detecting and manipulating periodic patterns in images, with applications in texture analysis and image restoration.
View Project →
Project 3: Face Morphing and Population Modeling
Developed a comprehensive face morphing system using Delaunay triangulation and affine transformations. Implemented features including mid-way face computation, smooth morph sequences between faces, population mean face generation, and caricature creation through geometric extrapolation. Created an interactive tool for facial feature manipulation and gender transformation effects.
View Project →
Project 4: Image Warping and Mosaicing
Developed an advanced image stitching system with both manual and automatic approaches. Implemented homography computation, image rectification, and panorama creation with alpha-gradient blending. Built a complete feature matching pipeline using Harris corner detection, ANMS, feature descriptors, and RANSAC for robust automatic panorama creation. Demonstrated the system on various indoor and outdoor scenes.
View Project →
Project 5: Fun with Diffusion Models
Implemented and deployed various diffusion model applications using Stability AI's DeepFloyd Model. Created a comprehensive pipeline for image manipulation including forward and reverse diffusion processes, inpainting with masks, text-conditioned image translation, visual anagrams (optical illusions), and hybrid image generation using factorized diffusion. Also implemented a UNet model from scratch for single-step denoising.
View Project →Final Project: Image Quilting
Implemented an advanced texture synthesis algorithm using image quilting and graph cuts. Developed methods for finding optimal patch boundaries, seamless texture transfer, and texture preservation. Created applications for texture expansion, texture transfer, and artistic image stylization using quilting techniques.
View Project →Final Project: Neural Radiance Fields (NeRF)
Implemented a complete NeRF pipeline for novel view synthesis of complex 3D scenes. Developed neural network architecture for learning volumetric scene representation, implemented positional encoding, hierarchical sampling, and volumetric rendering. Created tools for camera pose optimization and view interpolation.
View Project →