Picstonic Website
The Django half of Picstonic — a browser-based background remover and the processing API behind the Flutter app. No accounts, no dashboard: a visitor uploads an image and gets the cut-out back. Segmentation runs on classical OpenCV computer vision. Internal R&D at Bixosoft, never productionised.
What Problem Did This Project Solve?
One processing engine serving both a web page and a mobile app, with no account system to hide behind.
- The same segmentation had to serve two very different consumers — a browser form and a Flutter client — without maintaining two implementations.
- There are no accounts. A visitor arrives, uploads and leaves, so there is no session to attach work to and no user to hold state for.
- Background removal is CPU-bound. Several simultaneous uploads will saturate a server that naively processes each request inline.
- Uploaded images are user photographs that should not accumulate on disk indefinitely once the result has been delivered.
- Classical computer vision has real accuracy limits on fine edges and low-contrast subjects, and no amount of engineering around it changes that.
- The output needs a correct alpha channel, which constrains the encoding and delivery path — a JPEG round trip would silently destroy the transparency.
What Solution Was Delivered?
A single Django processing service with two thin front doors onto it.
- One segmentation pipeline in Python, exposed both through a Django-rendered upload page and through the API the Flutter app calls.
- Deliberately account-free: upload, process, return. No sign-up stands between a visitor and the thing they came to do.
- Classical OpenCV processing — masking, thresholding and morphological cleanup — producing a transparent PNG rather than a flattened image.
- PNG-with-alpha preserved end to end, so transparency survives from the processing step through to download.
- Uploaded files treated as transient and cleaned up after delivery rather than accumulating on disk.
- Server-rendered Django templates for the upload and result pages, keeping the web surface as light as the job it fronts.
What Can This Project Do?
Upload & Process
A visitor uploads an image and receives the background-removed version back, with no account required.
OpenCV Segmentation
Classical computer vision — masking, thresholding and morphological cleanup — rather than a learned model.
Shared Processing API
The same pipeline serves the Flutter app, so there is one implementation rather than two.
Alpha Channel Output
Results are delivered as transparent PNGs, with the alpha channel preserved end to end.
Transient File Handling
Uploaded photographs are treated as temporary and cleared after the result is delivered.
Django Service
Server-rendered upload and result pages over a Python processing core.
Built With Modern Engineering Standards
Have a project requiring modern web architecture, mobile apps, or autonomous AI pipelines?
Let's discuss how we can engineer a custom solution tailored to your operational bottlenecks.