This repository contains various cv projects. Below is a brief description of each project:
This project implements a digit recognition system using a Convolutional Neural Network (CNN). It includes a graphical user interface (GUI) where users can draw a digit, and the system predicts the drawn digit.
digit_writer.py: Tkinter application for drawing digits and predicting them from the network.digit_recognition.pt: Pre-trained CNN model for digit recognition.MNIST_CNN.ipynb: Jupyter notebook with code to develope the digit recognition model and train the network.handwritten_digit.png: Place holder image for images created from the gui.
This folder contains misc projects based on computer vision deep learning such as cellular cancer detection and transfer learing for image classification.
cancer_detection.ipynb: Jupyter notebook containing a self implemented U-Net CNN architecture for cancer imaging.transfer_learning.ipynb: Jupyter notebook containing a transfer learning example for image classification on the intel dataset with unfreezing implemented.
This project implements graph-cut segmentation, a technique for image segmentation. It provides a clean and finalized implementation of the algorithm.
segmentation.ipynb: Jupyter notebook containing graph-cut algorithm.gui_seg.py: Tkiner application for chosing foreground and background pixels to calculate pixel distribution.provided_images: Folder of images used for segmentaiton with their associated ground truth mask.
This project implements the Hough Transform algorithm for detecting circles in images. It includes a personally implemented version of the algorithm.
hough_transform.ipynb: Jupyter notebook with hough transform implementation for various images. Canny edge detector is also implemented here.Images: Various images for extracting gradient magnitude from as well as cirlce detection examples.
This project implements an image stitching algorithm to combine multiple overlapping images into a panoramic image. It includes fixes for output errors.
image_stitching.ipynb: Jupyter notebook containing image mosaic implementation based on OpenCV with SIFT and Ransac.Images: Images to be stitched together.