pipe leak recognition app Used AI

So the model and the whole app is now hosted on hugging face spaces, someone on the slack suggested this and it was a great fit for my app, everything is working fine now, including the video inference.

got a working version of the model and the web interface deployed on nest, check it out at https://flowingvision.leonardocerv.hackclub.app/

the model inference isnt the best bc open vino is optimized for intel and nest runs on amd, nevertheless i think i did a good job and ill be ready to ship too!!

made the web interface prettier, and focused on making the model actually run the inference live on camera, it was such a struggle but now it works, next step is making the user able to upload a picture

Hey guys so ive been trying to make a web interface for the ai model, but ive ran into some issues, im still gonna share the progress of what i have as for right now, the ultimate goal is for the model to be able to run on your camera without you needing to install anything.

DETECTION MODEL RUNNING! 💧

TL;DR: My AI leak detection model (made from scratch) is ready for anyone to check out, you can test it cloning my repo for this project, ill focus on adding a readme so the setup is a breeze for anyone. Feel free to hmu regarding issues or questions.

Hey people, this is my first of many devlogs for this project, i just finished a MASSIVE milestone in my Flowing Vision project and I'm REALLY EXCITED about the progress!

Improvements summary:
AI MODEL:
✅ Successfully implemented the leak detection system using OpenVINO for a faster inference
✅ Created a smooth pipeline from PyTorch -> ONNX -> OpenVINO using python programs
✅ The model now reliably detects water leaks with confidence tracking over multiple frames
✅ Added a wack confirmation logic that requires 20+ consecutive frame detections for a leak to be valid

DATABASE:
✅ Connected to MongoDB for persistent leak storage
✅ The model can still run without the need for a mongo connection
✅ Automatic database recording of confirmed leaks
✅ Base64 encoding for image (ss) storage
✅ Each leak gets a unique timestamp-based ID

VISUALIZATION:
✅ Implemented live camera feed processing (its low fps sometimes)
✅ Added real-time bounding box visualization
✅ Created on-screen status indicators for DB connection
✅ Added frame counters for continuous leak detection

WHAT'S NEXT?
I'll now focus on creating a better README with detailed setup instructions for the model testing. This will make it easier for anyone (including me) to get the system up and running quickly.

After that, I'm finally diving into building the Django app to create a proper web interface for this system. Time to move on to a web app that can display leak data, historical records, and the leak screenshots!

💧💧 Any feedback is very much appreciated! 💧💧