Pattern Studio (MS Capstone)

• Developed an Android application to apply a selected pattern on a clothing segment in a given user image.
• A deep learning-based model (UNET model) was used for instance segmentation to segment out the clothing on the backend server.
• Blended the pattern into the image using a linear blend operator.

Technologies Used: Java, Python, Pytorch, OpenCV, Flask, Docker
Frontend Code | Backend Code

Image Retrieval using Image Matching and Captioning

• Performed query-based search on a dataset of unlabeled data where Image similarity indexes were generated for all the images using CNNs with keypoint matching.
• Captions generated by an LSTM for random images were matched with the query and all the similar images of that matched random image were returned.

Technologies Used: Python, OpenCV, Pytorch

Code | Report

Design Pattern Generation using Generative Adversarial Networks

Used a Self Attention GAN on self curated clothing design pattern datasets to generate new and appealing patterns which were visually pleasing.

Technologies Used: Python, OpenCV, Pytorch

Slides | Report
Generated Results: