Industries / Media, arts and recreation / Museums
Conservation
Studying what objects are made of and how they decay, and slowing or repairing the damage.
6
Milestones achieved, from 1888 to 2025.
1
To watch for: announced, not yet confirmed outside.
2
Open deliverables. Nobody has shipped them.
The road
AchievedWatch forOpen
Achieved
1888
A laboratory in the museum
On 1 April 1888 the Royal Museums of Berlin founded a "Chemistry Laboratory", now the Rathgen-Forschungslabor. Its first director, Friedrich Rathgen, held the post for nearly 40 years; "his primary task was to improve the methods employed for the conservation of museum objects", and he also analysed materials and studied how objects were made.
Credit
Friedrich Rathgen; Royal Museums of Berlin
1920
Damage from wartime storage
In 1919 the chemist Alexander Scott looked into British Museum objects "alleged to have suffered deterioration from storage in the London Underground during World War I". In 1920 he became the first director of the museum's research laboratory, and Harold Plenderleith joined him in 1924.
LessonMoving objects to safety can harm them too, so emergency storage needs watching.
Credit
Alexander Scott; British Museum
1966
The Arno floods Florence
At 2.30 am on 4 November 1966 the Arno overflowed. In the church of Santa Croce, Cimabue's Crucifix was soaked in water, naphtha and mud, and 60% of its painted surface was lost, by the Uffizi's account. Many works were restored at the workshop of the Opificio delle Pietre Dure, with help from volunteers who came from all over the world.
LessonOne flood can damage much of a city's art at once, so museums need emergency plans and trained hands ready.
Credit
Opificio delle Pietre Dure; Uffizi Galleries
2008
A hidden portrait revealed
A team including Joris Dik of Delft University of Technology and Koen Janssens of the University of Antwerp scanned Van Gogh's Patch of Grass (Paris, 1887, Kröller-Müller Museum) with X-rays from the synchrotron at DESY in Hamburg. Mapping where mercury and antimony lay in the paint showed a woman's head painted underneath.
Credit
Joris Dik; Koen Janssens; DESY; Kröller-Müller Museum
2014
Wider climate bands
At meetings in Hong Kong and Melbourne in September 2014, the International Institute for Conservation (IIC) and ICOM's conservation committee (ICOM-CC) agreed that collection care "should be achieved in a way that does not assume air conditioning" and that museums should cut their energy use. They endorsed the Bizot Group's guideline for many materials: relative humidity (RH, how damp the air is) of 40–60% and a temperature of 16–25°C, changing by no more than 10% RH in 24 hours. The Bizot Group of museum directors published its Green Protocol in 2015; its 2023 handbook says "adopting broader environmental parameters does not compromise the safety of most collections", and that the new numbers should not simply replace the old rule of 50 ± 5% RH and 21 ± 2°C.
LessonA tight rule kept for decades can rest on habit, and loosening it can save energy without harming most objects.
Credit
IIC; ICOM-CC; Bizot Group
2025
Conservation in full view
The Rijksmuseum's Operation Night Watch, which it calls "the largest research and restoration project ever carried out" on Rembrandt's The Night Watch, is done "in a glass chamber, in full view of visitors"; the museum says the team began removing the varnish during 2025. The Art Newspaper described it in September 2026 as a five-year project.
Credit
Rijksmuseum
Watch for
A printed mask for losses
Alex Kachkine of MIT reported in Nature that software found 5,612 damaged areas on a 15th-century oil painting and filled them with 57,314 colours, printed on thin polymer films laid over the paint. The whole process took 3.5 hours, which he estimates is about 66 times faster than traditional methods, and the films "can be easily dissolved with conservation-grade solutions". MIT's report names no museum that has used it.
Achieved whena museum uses the method on a work in its own collection and conservators outside MIT publish an assessment of the result and of its removal.
Claimed by
Alex Kachkine, MIT, June 2025
Open
01
A reader for insect traps
A phone app that photographs a used sticky trap, finds every insect on it, names the museum pests (webbing clothes moth, carpet beetles and others), says whether each is an adult or a larva, and adds the counts to the room's record and floor plan.
The briefCredit · unclaimed02
Corrosion grades from any lab
A tool that grades the metal strips (coupons) from an Oddy test, from a standard photo taken in any lab. The Oddy test seals a sample of fabric, paint or glue with silver, copper and lead coupons at 60°C for 28 days, to show whether it gives off gases that would corrode objects in a case or crate; the result is a rating of permanent use, temporary use or unsuitable.
The briefCredit · unclaimed
Open 01 · the brief
A reader for insect traps
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A phone app that photographs a used sticky trap, finds every insect on it, names the museum pests (webbing clothes moth, carpet beetles and others), says whether each is an adult or a larva, and adds the counts to the room's record and floor plan.
Why it is still open
English Heritage's guideline says traps "should be checked four times a year" and "Insect pests caught on traps are to be identified and recorded", with doubtful ones sent to its pest manager or an entomologist; its staff log results on spreadsheets, and its catch data since 1997 show webbing clothes moth activity increasing. Museumpests.net lists "Identifying and counting invertebrates caught in each trap" as a routine task. Phone-photo methods already sort insects on yellow sticky traps in greenhouses (Böckmann and colleagues reported telling whiteflies from their natural enemies more than 85% of the time), but no published model or photo set for museum pests on museum traps was found.
english-heritage.org.ukenglish-heritage.org.ukmuseumpests.netpmc.ncbi.nlm.nih.gov
Built with what exists
A phone with a clip-on macro lens and a ring light; an open object detector such as Ultralytics YOLO; a small open vision model retrained on one graphics chip (see the Note "One GPU to fine-tune"); identification keys from museum pest guides; the spreadsheets that pest programmes already keep.
First step
Ask one museum to keep its traps from the next quarterly check, photograph each under the same light, and have the person who checks them label every insect.
Done looks like
On 300 used traps from at least three museums, labelled by an entomologist, the app's count of each species is within one insect of the expert's on at least 90% of traps, and it names at least 90% of the pests correctly. The labelled photos are published under an open licence.
Open 02 · the brief
Corrosion grades from any lab
The first to publish is credited on this line, and the deliverable moves from Open to Achieved.
What to build
A tool that grades the metal strips (coupons) from an Oddy test, from a standard photo taken in any lab. The Oddy test seals a sample of fabric, paint or glue with silver, copper and lead coupons at 60°C for 28 days, to show whether it gives off gases that would corrode objects in a case or crate; the result is a rating of permanent use, temporary use or unsuitable.
Why it is still open
Coupons are still rated by eye under strong light with a magnifier, and the American Institute for Conservation's wiki says "Oddy Test results are subjective" and that standardizing between testers and labs "is quite difficult". In 2022 Emily Long and colleagues at University College London trained a neural network on the Metropolitan Museum's coupon photos that classified 98% of a Met test set correctly, by their own test. Photos from other labs were of low quality, lacked ratings and drew "a high level of disagreement" among experts, so they were not used to train it; the paper sets out the steps still needed for "a universal" tool.
conservation-wiki.comconservation-wiki.comdiscovery.ucl.ac.uk
Built with what exists
A light box with a colour card for standard photos; the 2022 approach and the Met's labelled images as a starting point; a small open vision model retrained on one graphics chip (see the Note "One GPU to fine-tune"); coupons from the round-robin tests in which several labs run the same materials.
First step
Write a one-page photo protocol and ask three labs to photograph their next batch of coupons with it.
Done looks like
On 300 coupons from at least four labs, each also rated by three conservators, the tool agrees with the conservators' majority at least as often as the conservators agree with one another, and the photos, ratings and model are published.