5 ways to photograph a floor plan for the best AI results
Simple tips for capturing a sketch or printed floor plan so AI extraction reads your walls, doors, and dimensions accurately the first time.
On this page
- 1. Shoot straight down
- 2. Fill the frame, keep it sharp
- 3. Even, glare-free lighting
- 4. Include the dimension labels
- 5. One room, one plan
- Frequently asked
- Can I photograph my plan from a phone or monitor screen?
- My plan has handwritten notes and furniture sketched on it. Is that a problem?
- Do I need the whole home in one shot?
- Keep reading
AI floor-plan extraction is only as good as the image you feed it. A few seconds of care when you photograph a plan saves you corrections later. Here are five habits that make a real difference.
1. Shoot straight down
Hold the camera parallel to the paper, directly above the plan. Perspective skew makes straight walls look like trapezoids, which throws off both the geometry and the scale. If your phone has a document-scan mode, use it. It flattens the page for you.
2. Fill the frame, keep it sharp
Get close enough that the plan fills most of the frame, but not so close that edges blur. More detail on the walls and labels means more accurate extraction. Tap to focus before you shoot.
3. Even, glare-free lighting
Diffuse daylight beats a harsh overhead lamp. Glare washes out line work and dimension text; shadows hide it. Turn the page or move to a window rather than fighting reflections.
4. Include the dimension labels
If your plan has written measurements ("4.2 m", "3800", "12'6""), make sure they are in frame and legible. Kansumi uses them to calibrate scale automatically, and you will still confirm one, so a readable number means fewer taps.
5. One room, one plan
Upload a single floor plan per project rather than a multi-page architectural set. If you have a PDF, export just the page with the room you are designing. It keeps extraction focused and fast.
Ready to try it? Upload a floor plan and see the extracted layout in seconds.
Frequently asked
Can I photograph my plan from a phone or monitor screen?
You can, but paper beats pixels. Photographing a screen introduces moiré patterns and glare that look like extra lines to a vision model, and both lower extraction confidence. If the plan only exists digitally, export or download it as a PDF or PNG and upload that file directly. Kansumi accepts documents as well as photos, and a clean export always outperforms a photo of a screen.
My plan has handwritten notes and furniture sketched on it. Is that a problem?
Usually not. Extraction focuses on the strong, continuous lines that form walls and the symbols for doors and windows, and it assigns every detected element a confidence score, so a stray "IKEA?" scribble tends to be ignored rather than misread. That said, heavy markings directly on top of a wall can drag its confidence down. If a wall comes back amber instead of green, a quick crop or a cleaner re-photograph of that corner is the fastest fix.
Do I need the whole home in one shot?
No: one room is enough, and it's often better. Kansumi designs one room at a time, so a flat, frame-filling shot of the bedroom you actually want to redesign gives the extractor more pixels per wall than a whole-apartment plan shot from across the table. If you do upload a full floor plan, you'll simply pick the room to design during validation.