Working under Proff. Min Xu

Developed an image analysis pipeline that utilizes TensorFlow's EfficientNetB0 model to analyse and classify images. Implemented image preprocessing using Python libraries like PIL and NumPy, resizing images and converting them into a normalized array for deep learning model input. Built a custom image feature extraction function that converts processed images into vectors for further analysis. Designed a system to generate JSON reports containing image paths, feature vectors, and model prediction results.

Researched and developed American sign language recognition models using Convolutional Neural Network, LSTM and Recurrent Neural Network for real time video detection.

Headed Development of Detecting diseases in Groundnut Plants using Convolution Neural Network. Attained an accuracy of 73% on trained model. Preprocessed over 7,000 images by resizing, normalizing, and converting to RGB format for training. Achieved a test accuracy of over 80%. Evaluated the model’s performance using confusion matrices, classification reports, and multi-class ROC-AUC curves.

Developed an automated resume screening tool using StreamLit for the frontend and open source zero shot classification feature of hugging face model for backend predictions.