Upload any architectural floorplan and get back fully labelled rooms, detected doors and windows, clean SVG overlays, and a structured JSON export — in seconds.
Six processing stages, fully automated — from raw image to production-ready data.
SegFormer-based deep learning model detects wall pixels with multi-scale tiling for large plans. Output is a crisp binary mask.
YOLOv8 two-stage pipeline: bounding boxes first, then keypoints. Supports single-hinge, double-hinge, and windows.
Flood-fill the sealed wall mask, then simplify each room into a clean polygon snapped to dominant wall angles.
Tesseract OCR reads room names and callout numbers directly off the floorplan and assigns them to the correct room polygon.
No keypoint model? No problem. The built-in CV analyser uses adaptive thresholding and Hough circles to find hinge, latch, and arc-tip geometrically.
Every detection layer is exported as a transparent SVG overlay. A single pipeline.json carries rooms, detections, and usage metrics.
The pipeline handles everything automatically. You just provide the image.
Drop in a JPEG, PNG, or PDF. The pipeline auto-scales and tiles large plans so resolution is never a bottleneck.
Six stages run in sequence — walls, doors, gap sealing, room geometry, OCR, and association — typically completing in under 30 seconds.
Open the web editor to inspect, correct, and sign off on every room and detection. Export structured JSON or SVG layers for your workflow.
Start free. Upgrade when you're ready to run the pipeline.
Each stage produces its own visual layer, stackable in the editor.
Raw JPEG/PNG/PDF as uploaded
SegFormer wall mask overlaid on the plan
Bounding boxes, keypoints, and swing arcs
Flood-fill polygons with OCR room names
Add your own screenshots: replace the placeholder divs in the gallery cards with
<img src="images/your-screenshot.jpg" alt="...">
A transparent look at what's been built and what's coming next.
SegFormer wall model, YOLOv8 two-stage detector, flood-fill room geometry, Tesseract OCR labelling, and a complete SVG + JSON output pipeline.
Flask app with live SSE log streaming, per-project SVG layer editor, room re-detection, sign placement, and export tools. User auth via SQLite + bcrypt.
Pure computer-vision alternative to the ML keypoint model. Uses adaptive thresholding and HoughCircles to locate the hinge, latch, and arc tip without any model weights. Includes per-crop debug output and geo-kp SVG overlays.
Geometric keypoint support for double-hinged doors. Native PDF vector extraction to sharpen wall masks for high-DPI CAD exports (wall_vector_snap.py promoted to default-on).
REST API with API keys, per-call billing hooks, and batch job queue (Celery + Redis). Webhook callbacks for completed jobs. Multi-tenant project namespacing.
AWS deployment (ECS/Lambda) with per-job cost metering. BIM/IFC export so rooms and openings can feed directly into architectural and FM workflows.
Create an account or sign in to access the editor.
Create your free Planqix account. Use the full registration page or fill in the quick form below.
We'll be in touch at the email address you provided.
Already have an account? Sign in to open Planqix.
You'll be redirected to the login page if not already signed in.