How this site works
The interactive parts of this site, and what each one does with your data.
Photo and model output
The opening portrait has a Photo / Model output switch. Model output shows the boxes, masks and pose that YOLO11x-seg and YOLO11x-pose produced when they ran once on this photo, offline. The page stores those results as data, runs no model, and keeps the mistakes visible. Switching needs no JavaScript.
Mural restoration slider
Under Papers, a slider compares a damaged Rajasthani mural with the restored output from the CVIP 2023 paper. Damage mask shows the regions the model marked as damaged. Arrow keys, touch and screen readers can all move the slider, because it is a standard range input.
ask_soham chat
The chat sends your question to a server function, which asks a model hosted on Groq to answer as me. The model reads only the content this page is built from, plus a few notes, and the API key stays on the server. It can be wrong, so email me for anything that matters. The Ask about this buttons on each job and project open the chat with a question filled in.
Email follow-up
After the chat answers, you can ask me to reply by email. If you tick the box and send, your address and the conversation go to my inbox through Resend, an email service, and I reply from my own address. This site saves neither. Resend keeps its usual delivery log.
llms.txt
A plain-text file for AI agents: the facts on this page, plus notes on how to work with me and what not to claim. Give it to an agent and ask about me.
Local editor
I edit this text in place on a copy of the site that runs only on my laptop, and leave myself comments there. The public site ships none of that editor’s code.