Sanyam Jain

I am a PhD student in Generative AI at the Department of Dentistry and Oral Health, Aarhus University, Denmark (2024 - present), working with Prof. Ruben Pauwels, Prof. Alexandros Iosifidis, and Dr. Bruna Neves de Freitas. My PhD project is on Synthetic Dental Radiography using Generative AI.

Previously, I completed my MSc in Applied Computer Science (2022 - 2024) at Ostfold University College, Norway, advised by Prof. Stefano Nichele, where I worked on Artificial Life, Cellular Automata, and emergent complexity in Lenia. I was also a Summer Intern at NTNU Gjovik (2024) and OsloMet (2023). Earlier, I was a Junior Research Fellow at IIT Jodhpur (2020 - 2022), part of the VANETs Lab and the QIC group, and I received my BTech in Computer Science from UPES Dehradun (2015 - 2019).

My research interests include Representation Learning, Generative Models, Computer Vision, and Medical Imaging, with current focus on diffusion models for dental radiography.

Email  /  CV  /  LinkedIn  /  GitHub  /  Google Scholar  /  Twitter

Water Your Repo, Grow Your Greens!

Proposes a two-stage pipeline combining a latent diffusion model with super-resolution to synthesize high-resolution panoramic dental radiographs, addressing data scarcity in dental imaging while preserving anatomical fidelity.

Develops a generative adversarial network for synthesizing panoramic dental radiographs, evaluated for clinical realism by dental experts. The work demonstrates the potential of GAN-based synthesis for augmenting limited medical imaging datasets.

Investigates underwater object detection on Denmark's Brackish dataset using a modified EfficientDet. The proposed approach achieves 88.54% IoU with five-fold cross-validation, outperforming prior baselines on the dataset.

Studies the robustness of YOLOv5 on traffic sign detection under a range of adversarial attacks including L-BFGS, FGSM, C&W, BIM, PGD, One Pixel Attack, and Universal Adversarial Perturbations, and analyzes their effect on detection accuracy.

Explores emergent complexity in Lenia, a continuous cellular automaton, using evolutionary fitness measures based on Variance over Time (VoT), Autoencoder reconstruction (AE), and a combined AEVoT criterion to discover novel life-like patterns.

Working on Synthetic Dental Radiography using Generative AI - building diffusion and GAN-based models for panoramic dental radiograph synthesis, super-resolution, and downstream clinical evaluation.

Research internship in the Education Technology Lab on Computer Vision and Large Vision Models for educational applications.

Continued research on artificial life, evolutionary computation, and emergent complexity in continuous cellular automata (Lenia) and Neural CA.

MSc research on complexity in discrete and continuous cellular automata using evolutionary algorithms; led to multiple publications on Lenia and underwater object detection.

Research in Machine Learning, Deep Learning, and Dependable AI. Teaching Assistant for Machine Learning (Winter 2022) and Foundations of Quantum Information and Computation (Trimester 3, 2021).

Interactive portal for exploring emergent creatures discovered in the Lenia cellular automaton via evolutionary search. Built with HTML/JS to visualize evolution runs, fitness landscapes, and discovered patterns.

Project page for DeepSeaNet, presenting underwater object detection results on Denmark's Brackish dataset using EfficientDet with cross-validation and qualitative comparisons across baselines.

Studies complexity and emergent behaviour in elementary CA, discrete CA, and 2D Multi-Neighborhood Cellular Automata, with a frequency histogram based coarse-graining method to characterize pattern formation.

Implementation of Grad-CAM visualizations for YOLOv5 and analysis of how several adversarial attacks (FGSM, PGD, BIM, C&W, One Pixel, UAP) shift the network's attention on traffic and road sign detection.