Image Sorter
Image Sorter
An Image gallery application in developed in Python with Tkinter frontend that offers advanced Image processing and AI features. The application uses both conventional computer vision algorithms, OpenCV modules and Deep learning models to generate its functionalities
Skills developed:
Software: Process based parallelism for multi-processing, Frontend development with python, OpenCV functions and methods
Algorithms: Dynamic programming for stereo vision and depth estimation, Homography, basic convolutions, numpy operations, K-means, Hierarchical Agglomerative Clustering
Deep Learning: Convolutional Neural Networks (VGG)
Features
Graphical User Interface
Similar Images
Hierarchical Agglomerative clustering used for finding the most similar images and sort them out in separate folders. Some results are shown below. Test accuracy 82 %
Stylizing Images
Convolution with filter operations performed to produce stylized images
Face Detection
Dlib library used to detect images of people in the album
Additional algorithms learned in the Coursework
Stereo Vision and depth estimation
Dynamic programming to calculate the depth of objects in an image if a pair of rectified images are provided. Results are shown below
Two images from the left are the pair of rectified images and the image at the right side is the depth estimated image. Higher the brightness of the pixels, closer the objects are. The red pixels are the occlusions.
Segmentation
K means for segmentation task. Results shown below