Deep Learning Assisted Autonomous Navigation over Economical Hardware: Implemented a cost-effective autonomous driving system on NVIDIA Jetson Nano. Utilized SSD-MobileNet for object detection and FCN-ResNet34 for road segmentation.

Developed a model using the FER-2013 dataset to classify emotions. Achieved optimal performance with CLIP ViT-B/32, balancing accuracy and speed for real-time applications.

Developed a real-time fire detection system by fine-tuning YOLOv8n on a specialized dataset. Trained on a P100 GPU with 1004 training images, achieving optimal balance between speed and accuracy. Created a Python application for video inference to identify fires in real-time footage.

Created a Streamlit app for pneumonia detection using ResNet18 with 80% test accuracy. Deployed on Hugging Face Spaces for real-world accessibility.

Trained models to classify happy, sad, and surprise emotions using MediaPipe-extracted landmarks. Achieved 77.7% accuracy with Logistic Regression.

Content Based Image Retrieval: Built a system using ORB features and FAISS indexing to retrieve similar images. Processed datasets of cars, cricket balls, and bats for accurate query matching.

Developed a Streamlit app to generate summaries of YouTube transcripts via Gemini API. Participated in the Gemini API Developer Competition.

Image search: React app for searching images via Unsplash API.

Country Information: Get Country Information by Country Name.