OpenDerm: Open-source skin imaging under $8,500

Skin imaging for melanoma checks can use wide-field total-body photography or guided smartphone imaging, and each gives something up. The site says many wide-field systems lose fine detail, and it lists smartphone detail as variable. OpenDerm, an open-source robot designed and built by Marion Lepert, takes a different route: it moves one camera close to the skin and builds a 3D map from many sharp photographs.

The project publishes a capture resolution of 78 pixels per millimeter, a hardware cost under $8,500, and open licenses for both software and hardware. It is a research tool and not a medical device.

What OpenDerm is: an open-source robot for skin imaging

OpenDerm is a robotic imaging system that photographs skin at close range and combines the photos into a high-resolution 3D map of the skin. Marion Lepert designed and built it, and the hardware designs, software, and build documentation are openly available.

OpenDerm is imaging hardware, not a skin analysis app. It takes pictures and builds maps, and the project site describes no model that interprets the images.

The map is meant to be repeatable. Images taken at different times and aligned to each other (the site calls these registered scans) help reveal new lesions and subtle changes in existing ones. The site also marks the project as not a medical device and for research use only.

Why early melanoma detection depends on better imaging

The melanoma figures the project cites

The project starts from melanoma, a form of skin cancer, and reports that more than 100,000 Americans are diagnosed each year. Five-year relative survival is nearly 100% when the cancer is found early, but about 34% once it has reached distant parts of the body. The site's first footnote cites the National Cancer Institute's SEER Cancer Stat Facts: Melanoma of the Skin. The site adds that about 30% of melanomas develop within an existing mole while 70% appear as new lesions.

Three areas that need progress

Total-body photography (photographing the whole skin surface) can provide a record of the skin over time, but access stays very limited. The site gives two reasons: high costs, and limited clinical evidence that existing systems reliably catch melanoma earlier than dermatologists. The project names three areas where total-body imaging needs to improve:

• Higher-resolution imaging. Many existing total-body photography systems use wide fields of view to cover large areas of skin, at the expense of fine detail.
• Longitudinal datasets (images of the same skin repeated over time). Most AI models for skin cancer detection are trained on isolated images of lesions already flagged as suspicious, and those datasets show little of how melanomas first emerge and evolve.
• Frequent, accessible scanning. Total-body photography has to become more affordable and widely available, and patients at high risk should be able to get frequent, standardized scans.

The project says a robotic imaging system offers a practical way to meet these requirements. A camera that moves in close and follows the body's contours can keep distance, angle, focus, and lighting consistent, and one camera on cheap actuators costs less than a large array of fixed cameras. The project's design is public, and its code is on GitHub.

How robotic scanning compares with total-body photography

Three approaches compared

The site sorts imaging of large skin areas into three approaches, each trading speed, detail, consistency, footprint, and cost differently. The three-approaches table gives the speed, detail, and cost labels it assigns and the examples it lists.

Wide-field systems photograph the body from a distance, and the site says many existing ones give up fine detail to cover large areas of skin. Guided smartphone imaging needs no dedicated hardware, but the site lists its speed as manual and its detail as variable.

Close-range scanning trades speed for detail

Close-range robotic scanning puts a high-resolution camera on a robotic arm or a gantry (a frame that carries the camera along rails). The camera moves near the skin and works across the body region by region, holding its distance and viewing angle under control. One camera can then record much finer detail than a wide-field system standing back, and it needs less space and less camera hardware.

The cost is speed. The body is scanned in sequence instead of all at once, and the site lists close-range robotic scanners at minutes where wide-field systems take seconds. Total-body photography as a whole is also still an emerging field: clinical availability in the United States is limited, and many systems are experimental or in development.

How the robot builds a 3D skin map

OpenDerm turns 3D skin imaging into a sequence of steps: move to a station, set the camera, take a photograph, then merge all the photographs into one surface. The moving frame, the sensor head and the single camera each handle one part of that sequence.

Four axes built from off-the-shelf parts

OpenDerm is built mostly from off-the-shelf parts, meaning ready-made components rather than custom ones. The camera rides on a gantry that moves with four degrees of freedom. Three linear axes place the sensor head in the workspace (X along the side rails, Y across the top beam, Z for camera height), and a rotary axis called RX tilts the head so the camera faces the skin surface.

What the sensor head carries

The sensor head carries a Canon EOS R7 camera with an RF 100 mm macro lens, a Godox MF-R76 ring flash with a cross-polarization filter, and two downward-facing laser distance sensors. The filter reduces specular glare, the shiny reflection off the skin surface. The laser readings let the control system hold the working distance and camera orientation steady relative to the skin.

At each imaging station the robot uses the two laser sensors to set the camera angle and working distance, then takes a high-resolution photograph. Afterward, the pipeline aligns the overlapping photographs and rebuilds them as one continuous surface.

Why the design uses one moving camera

A robot that moves close to the body and follows its contours can keep distance, angle, focus, and lighting consistent while it captures images across the skin. A single camera on inexpensive actuators (the motors that move each axis) also costs less than a large array of fixed cameras. The price of that choice is the speed trade-off described above.

Resolution and cost: 78 pixels per millimeter, under $8,500

OpenDerm captures skin at 78 pixels per millimeter, meaning a 1 mm stretch of skin spans 78 pixels. At that density the site says the scans resolve a mole's pigment network (the fine pattern inside a mole) and the individual hairs next to it. The site shows this in an interactive viewer with two registered 3D scans, where lesions are measured and matched between them.

The hardware costs less than $8,500. The site says it provides some of the highest-resolution imaging available at a fraction of the cost of commercial total-body photography systems. It compares cost only by tier: $$$ for wide-field systems, $$ for close-range robotic scanners, and $ for guided smartphone imaging. It gives no dollar figure for commercial systems, so it never says how small that fraction is.

The detail comes with slower capture: the site lists a close-range scan at minutes, against seconds for wide-field systems. It also publishes a complete bill of materials for builders.

Why skin cancer detection AI needs scans of the same person over time

The project argues that skin cancer detection AI needs a different kind of image: repeated scans of the same skin over time. Most models are trained on isolated pictures of lesions already flagged as suspicious.

Such skin lesion datasets show little of how a melanoma first emerges and evolves before it draws clinical attention. Learning earlier signs takes repeated high-resolution images of the same skin over time, which the site calls a longitudinal dataset.

OpenDerm's registered, repeatable scans fit that need. A low-cost robot that photographs the same skin under the same conditions again and again is meant to make that record practical. The site also says patients at high risk should be able to get frequent, standardized scans.

OpenDerm itself is not an analysis model. The official site does not describe a trained model, a published dataset, or clinical validation, so OpenDerm is better read as a way to gather data for melanoma detection AI than as one.

Open licenses, build files, and the research-only limit

The software is under the MIT License, a permissive software license. The hardware design files, schematics, bill of materials, and build documentation are under the CERN Open Hardware Licence Version 2 - Permissive (CERN-OHL-P-2.0). The two licenses split by scope: MIT covers the code, and CERN-OHL-P-2.0 covers the physical design.

The site publishes the complete bill of materials, CAD assembly, electrical and safety schematics, and step-by-step build instructions, with the source on GitHub. So the parts list, the 3D design, the wiring, and the build steps are all public.

The footer marks OpenDerm as not a medical device and for research use only. Treat it as a research tool: its scans are not medical advice, and the figures above are reported as the site gives them. OpenDerm is an open design for skin imaging that trades scan speed for detail and low cost, and it stays within research use.