Aashwin Sharma

Sign2Text

• Led a team in an AI project that translates common phrases of the Indian Sign Language (ISL) to text in real-time.
• Extracted the key points of the hands, joints, and face using MediaPipe and OpenCV.
• Made the dataset from scratch and trained it on LSTM neural network for prediction of the signs.

Autonomous Pest Control using Neural Networks

• Built a working model to automatically spray pesticides when an agricultural pest is detected.
• Detected and classified 8 types of pests using EfficientNet B3 (CNN) architecture integrated with the hardware periphals Raspberry Pi 4 and servo motor.

Image Colorization 

• Created an end-to-end deep learning model using Conditional GAN that learned to convert a given black and white image to its colored version.
• The model was trained on the COCO dataset and it achieved excellent results at image generation tasks, producing real-looking images.

Cart Pole Balancing

• Simulated a cart pole environment using OpenAI gym and used Deep Reinforcement Learning to train the agent to balance itself when set free.

• Secured AIR 3822 in JEE Advanced 2020.
• Bronze Medal at the 10th Inter IIT Tech Meet hosted by IIT Kharagpur.
• Gold Medal at the 55th & 56th Inter IIT Sports Meet in Football.