Projects

Pokemon Diffusion
Designed an end-to-end diffusion model using a classifier free guidance approach for image generation of pokemon types.
Devikalyan Das*
: Python
Rendering Competition 2021/22:
Created a 3D renderer for Computer Graphics 1 course at Saarland University by integrating concepts which we learnt during the course period
Devikalyan Das*, Pranay Raj Kamuni*
: C++
Semantic Segmentation of Cityscapes dataset using Pytorch:
Carried out Semantic Segmentation of Cityscapes dataset with deep learning architectures using Pytorch:
Devikalyan Das*
: Python
Super Resolution using GAN:
Carried out a novel GAN based architecture for super resolution with wider activation channels, regularization in the network and a novel loss function based on LPIPS:
Devikalyan Das*, Shashank Agarwal*,Nobel Jacob Varghese
: Python
Optimizers in OOD generalization:
Learnt the impact of Learning Rate, Optimizer, Learning Rate Scheduler and Batch Size on OOD generalization in Deep learning
Devikalyan Das*, Pranay Raj Kamuni*, Saurabh Kumar Pandey
: Python
Realistic Image Synthesis:
Learnt the advanced concepts of computer Graphics such as Monte Carlo sampling, Path Tracing, Next event Estimation, Progressive Photon Mapping and HDR with Tone Mapping.
Akshay Mundra*, Devikalyan Das*
: C++, C#
Semi-supervised Image Classification:
Explored various techniques of semi-supervised learning(ssl) for image classfication task such as pseudo-labelling, virtual adversarial training and FixMatch.
Devikalyan Das*, Rushan Mukherjee*
: Python
Monitoring Traffic Control:
Using deep learning approach for estimating traffic density using real-time images captured from smartphone and deploying it using android application TrafficMode
Akshay Mundra*, Beste Ekmen*, Devikalyan Das*, Kaustubh Sakhalkar*
: Python, Java