Projects
Distributed Hash Table
- Implemented a Highly Scaleable P2P Distributed Hash Table.
- Used techniques to make it tolerant to peer churns.
- Used Chord as underlying algorithm. View Project
Image to Paragraph generation using Visual Attention
- Developed a model that could generate a relevant paragraph by using an image as context.
- Used Attention Model to understand different parts of the image.
- Combined it with a self trained Language Model to generate a meaningful paragraph. View Project
Low Light Image Enhancement
- Implemented a model that achieved >30 PSNR when trying to produce a bright and enhanced image from a low light image.
- Used a multi branch fusion model to apply enhancement via multiple subnets.
- Used a combined loss function that gave more emphasis to dark regions to reduce noise while enhancing the image. View Project
Predicting CPU usage in Cloud VMs
- Developed a model to preemptively calculate future CPU utilization with <1.1% error.
- Compared the performance of LSTM, GRU and Independent RNN models. View Project
Message dependent Image Steganography
- Implemented an advanced steganographic technique that is an improvement over traditional methods.
- Used payload to determine embedding positions on the image. View Project