IN-PIC-ANALYSER - Stardance

Ship #1 Pending review

@Mouad on IN-PIC-ANALYSER · about 2 hours ago

I built IN-PIC VISION, an interactive computer vision web app that detects objects in uploaded images using a YOLOv8 backend and visualizes the results with real-time bounding boxes, confidence tags, and detected object breakdowns.

The most challenging part was the deployment pipeline. Getting PyTorch, OpenCV, and all native C/system libraries working cleanly in production without hitting container memory limits or broken system dependencies was tough. I ended up hosting the backend on my own server, supervising it with PM2, and exposing it securely using Cloudflare Tunnels to handle the public traffic and SSL.

I’m proud of building a complete, responsive dark-mode UI from scratch and connecting it smoothly to an end-to-end inference pipeline that runs fast and works directly in the browser.

To test it: Click “Try project” (or open the link), drag and drop or select any everyday image (cars, people, street scenes, bikes), and watch it scan and highlight all detected classes in real-time.

1 devlog
2h

Try project → See source code →

Open comments for this post

@Mouad on IN-PIC-ANALYSER · about 3 hours ago

1h 48m 52s logged

yo so i basically built this object detection app called in-pic vision using yolov5 and flask, at first i tried making a desktop app with tkinter but mac was being super annoying with blank screens so i just scrapped that and turned it into a web server instead, now you can just drag and drop any photo under 5mb and it runs through the model in like a couple ms, highlights all the objects with colored boxes, counts them up in the sidebar, and i made the whole ui dark mode with a laser scan animation and added links to my github and stardance profile