Using New Tech to Investigate Old Photographs

In 1860, roughly 30 years after photography was invented, a series of photos were taken across Europe and India. Many of these were captured by prominent photographers at well-known landmarks and sites. But in other cases, it’s still unclear where, when, and by whom these photographs were taken. 

This album is one of hundreds in the possession of Amsterdam’s Rijksmuseum, the largest museum in the Netherlands. The Rijksmuseum has become a leader in the effort by galleries worldwide to digitise their collections. The aforementioned travel album, which can be viewed online here, contains 60 images. Most of these photographs were taken by well-known photographers such as Samuel Bourne, and the duo Shepherd & Robertson. However, the location and authors of 21 images remain unknown.

Today, open source research can provide new clues. New tools and a wealth of online data make it easier to navigate more than 150 years of evolving landscapes, cities, buildings and street names. Reverse image searches, Google Lens, digitised newspapers, heritage and auctioneering websites, AI colourisation and tools such as Peakvisor can help add valuable information and understanding to historic art collections.

These tools and methods allowed us to pinpoint the location of several of these photographs. Here’s how we did it — and how you could, too.

Determining Locations

The 21 photos all appear to be original, but in many cases a reverse image search can still be used to find similar looking images that might provide more detail about location and time. 

To use a reverse image search, the photograph in question must be the only image in the frame. Cropping may be required.

 

A photo in the album digitised by the Rijksmuseum, location unknown.

A Yandex reverse image search of this cropped photo returns recent results showing pictures from the Königssee, a lake in the Bavarian Alps.

The slider below compares the Rijksmuseum’s photo to an image uploaded to Tripadvisor in May 2018.


Another photo from the German side of the Alps can be matched to an old postcard. Despite the different position of the camera, and new buildings and church towers on the postcard, the mountain range in the background provides a distinct landmark that the Yandex reverse image search picks up on. The search engine found a postcard of Berchtesgarten from 1911, offered for sale online. The slider below compares the two.

Because both of these locations are near the Königssee, we can infer that the photo of the lake below was likely recorded there as well.

 

An album photo placed near the previous two, location unknown.

Interestingly, the three big reverse image search engines (Google Images, Yandex, and Bing) don’t immediately recognise this location. However, using Google Lens (also accessible through a browser) immediately finds the right spot on the Königssee.

Peakvisor, a web and app service which recognises landscapes, can be used to match the mountain ridges in the picture to this location.

The album photo overlaid with Peakvisor’s tool for identifying mountain tops.

Another photograph in the Rijksmuseum’s album shows a mountain village. A reverse image search of this photo leads to a matching vista in seen in a photo found on an auction website, which was sold in 2017. The same website names the photographer as Charles Soulier (1840-1875) and the location as Chamonix, France, at the foot of Mont Blanc.

Apart from its larger size, the auctioned photo looks remarkably similar. As can be seen in the slider below, the buildings all appear to match and the angle is only slightly different. The main difference is the level of snowfall visible on the mountain above the mountain above the glacier. This glacier’s name — the Glacier des Bossons — is also given in the website’s description of the photograph.

With a simple Google Image search, we can establish that the auction website has given the location correctly: this TripAdvisor photograph by a tourist shows the glacier as viewed from Chamonix.

The album also contains three photos of a Gothic church — one showing a view across a river, the second showing the church’s tower, and a third of the church’s main door.

Three photos from the Rijksmuseum album

Searching the first photo with Yandex, Bing, and Google Images does not lead to immediate matches. But Google Lens does identify several cities with large churches or cathedrals, even though the church clearly looks different from many of them.

A Google Lens Reverse Image Search connects an album photo to a view of Ulm.


The building is the tallest church in the world, still standing today in Ulm, Germany. The reason the church looks different is because a spire was added roughly 30 years after the photo in the album was taken.

Google Images, Yandex, and Bing could not find a match for the next photo, although Google Lens could.

We can assist the algorithms of the three search engines which couldn’t by colourising the photo.

 

An photo from the Rijksmuseum’s album

But the photo can be enhanced using AI colourisation, available free online using a site like Hotpot.ai, which enhanced the photo as seen below.

 

An album photo that has been colourised using AI.

A subsequent Yandex reverse image search using the colourised photo now returns a clear match leading to other photos showing the Swiss city of Bern.

A Yandex Reverse Image Search using the colourised album photo leads to other photos of Bern.

But even Google Lens has its limits. Together with the other reverse image search tools they do not immediately recognise the following photo, even with cropping:

 

A photo from the Rijksmuseum’s album, location unknown.

This photo was placed in the album alongside two other photos of a castle and a church taken in a valley. These, on the other hand, could be found through Google Lens.

They show the castle of Tarasp and the church at Scuol two towns in the canton of Grisons, in eastern Switzerland.

Photos of the Castle of Tarasp and the church in Scuol identified via Google Lens, placed adjacent to the photo with unknown location in the album.

By locating the previous two photos on a map, we can check the surroundings on Google Earth to help pinpoint the likely location of the third image.

The building turns out to be the Büvetta Drinking Hall at near Tarasp a structure was built in the 1870s for vistors enjoying the region’s famous mineral water.

 

The perspectives of three album photos mapped with Google Earth Pro.

Google Lens’ recognition algorithm has improved the potential of reverse image searches for finding a match for old photos. Combined with other tools, such as PeakVisor, Google Earth, and AI colourisation, it has now become easier to geolocate photos of unknown origin.

Determining Time

But when were these photographs taken?

If the author is unknown and no date is written on the photo, physical features can help determine the time period.

The size of photos, as well as the chemical process by which photos were produced, change over time. Sites including GraphicsAtlas.org can help establish the rough time period on this basis, though this is a nuanced process best left to experts. These photos from the Rijksmusem are already identified as albumen pictures, a method of photography popular between the 1850s and 1900s, and have a rough period assigned to them in the album. 

But the content of photos can also be used to more precisely establish when they were taken. 

Reverse image searches of the photo of this church facade, also seen in the view across the river which we found using Google Lens reverse image search, leads to a match with the website of Ulm’s Art Association. The photo, taken in 1854 and described as the oldest known photo of the church, is almost identical in perspective to the Rijksmuseum photo. 

However, despite the similarities we can tell that the Rijksmusem’s photo is more modern, because the building has features that do not yet exist in the oldest photo.

 

Left: Oldest photo of Ulm church tower in 1854. Right: Photo from the Rijksmuseum album showing structural changes to the church, circa 1860 – 1864.

Now that we know the picture is from Ulm, we can search manually through Ulm’s City Archive to find images that have not been indexed by search engines.

This eventually leads to the exact same picture, though the Ulm’s City Archive’s version is of poorer quality. The timeframe, however, is more specific than that offered by Rijksmuseum (1865-1875), putting it somewhere between 1860 and 1864. 

In cases where a building has not undergone significant changes, its surroundings can be used to help determine when a photo was taken.

For example, the photo below from the Rijksmuseum album can easily be traced to Freiburg in Germany through a reverse image search. The museum also dates it from 1865 to 1875.

The results include very similar, though not exact matches, to other photographs. One of these was taken by Gottlieb Theodor Hase, a German photographer who took a series of photos of the historic cathedral city.

One of Hase’s photos of Freiburg cathedral was auctioned on artnet. Its caption states that Hase’s photo was taken between 1850 and 1859.

Features of the roofs surrounding the cathedral differ in the two photos. In Hase’s photo, two large dormer windows can be seen in the roof of a house near the cathedral square. The Rijksmuseum’s photo shows two smaller windows, much like those of surrounding rooftops, in the same place.

Left: a photo in the Rijksmuseum album. Right: a photo by Gottfried Theodor Hase between 1850 and 1859. Highlights by Bellingcat.


Searching again for Hase’s name also leads to another photo of the Freiburg cathedral also digitised by the Rijksmuseum (though not contained in the aforementioned album of 60 images). This photo, which is damaged, is very similar to the other Rijksmuseum photo, including the two large dormer windows, but was taken from a different angle. The Rijksmuseum dates this photograph, also by Hase, from 1852 to 1863.

Photograph of the Freiburg cathedral by Hase, also in the Rijksmuseum collection, dated 1852-1863

Assuming that the house’s owners did not go to the expense of reverting these new dormer windows to their previous state, this means that photos showing the smaller windows on the same building must have been taken earlier. This allows us to infer the sequence in which the three photographs were taken.

This means that the two photographs digitised by the Rijksmuseum (one by Hase and the photograph by the unknown author) must have been taken after the photograph seen on the Artnet website.

While the dormer windows facing us (in the red circle) appear in both, the newly added dormer windows (in the blue circle in the image below) are missing from the photo by the anonymous author.

Comparison of dormer windows seen in two buildings in both photos of Freiburg cathedral which were digitised by the Rijksmuseum. Unattributed photo from the Rijksmuseum album (top), photo by Hase in the Rijksmuseum collection (bottom).

Thus even if we cannot establish a precise year, the absence or appearance of various features in each photograph allows us to at least verify the accuracy of the sequence of date ranges provided.

Determining the period in which a photo was recorded through this method depends on other footage captured of that scene around the same time, along with deductions from the context of the photograph. The earlier this is, the harder it gets, as the number of available photos dwindle, and become lower resolution.

Drawings and paintings can be of use too, though these can be less accurate. In such cases, other online sources can be used to pinpoint when a photo was recorded. The following photograph provides a useful case study.

The Case of (Mc)Conway Hart

One of the photos in the Rijksmuseum album shows a street described in the album as “broken up”.

An album photo, location initially unknown

This picture has no identified author, date, or location. It is placed in the middle of the album, between pictures taken in Chamonix, France, and in Delhi, India. It shows several buildings alongside a broken-up street, which are named Jewellery James Browne, Mountains Hotel, and Photostudio McConway Hart. An internet search for these names does not generate useful results.

The photo studio could offer a clue for the possible author(s) of this picture, but searching for combinations of McConway and Hart also yielded no useful results.

A church tower visible in the distance, on the right edge of the middle of the photo, provided a useful reference point for geolocating this photo.

 

The album photo, with the church tower circled.

This tower appears very similar to a church pictured in another photo, placed near the end of the album. The location of this other photo was already identified as showing the Tipu Sultan Shai mosque in Calcutta, India.

 

The same church tower is visible in two separate album photos.

A comparison of the structures in both photos reveals it is likely the same tower. 

A comparison of the church tower in both album photos.

Because both the mosque and the church are still standing today, the angle of the old photograph can be used to find the ‘broken-up’ street on Google Maps. The street is called Sido Kanhu Dahar, and it leads to a grand building which, before India’s independence, was called Government House and nowadays houses the Governor of West Bengal.

Now that the street is known, can we find out more about the photography studio?

 

Detail from photograph


Searching for the street name in connection with business names does not lead to useful results. But searching for the street itself brings us to a Wikipedia page, which explains that Sido Kanhu Dahar was previously called Esplanade Row East

A search for the historical street name together with “Hart” and “photography” leads us to a website about the history of photography which mentions a Conway Weston Hart, who ran his business from Esplanade Row 7, Calcutta. 

Conway Weston Hart was an Australian portrait painter and photographer. His works are in the Getty Museum in Los Angeles, US, the National Portrait Gallery in London, UK, and in Canberra, Australia, as well as the National Gallery of Victoria, Australia. Conway Hart’s art is discussed further in an Australian magazine Australiana, which notes that his clientele “included dignitaries, politicians, members of the judiciary and their wives”. According to auction websites, one of his portrait paintings was sold for about $125,000 in 2016.

But why does the business sign appear to say McConway Hart, rather than Conway Hart?

One may well mistake MCCONWAY for a Scottish surname with the prefix Mc-, complicating further searches. It is more likely that the ‘MC’ stands for Master Craftsman, a title which Conway could well have used as he was both a photographer and painter. It’s one of several abbreviations and acronyms for trades and crafts which have fallen out of use, but can still be found in historic documents.

According to the National Gallery of Victoria Hart left Australia in 1861 for Calcutta, together with his wife, to set up his studio there. Only 3 years after arriving in India, he died of cholera. India being a former colony, Conway W. Hart’s obituary can be found in The British Newspapers Archive online.

 

In The Albion of April 25, 1864, the obituary for Conway W. Hart notes that he died of cholera on March 3rd of that year.

Conway also had children of working age. Digitised newspapers, birth registries, and family genealogy sites such as MyHeritage.com and Ancestry.com, reveal that the couple had at least three children, the oldest son being 15 at the time of their father’s death.

Nonetheless, Conway’s photography business did not appear to last after his death. Searching archived newspapers using the business address leads us to a real estate advertisement. This ad was posted on June 22, 1865, over a year after Conway’s death, and it contains Conway’s business address, which at this point is being used by a real estate agent.

 

An advertisement from June 1865, using Conway Hart’s business address.

The image below, found on the website Old Indian Photos, shows that the building’s facade had changed by 1865. By zooming in and comparing it to the original photo, we can see that the building that used to house Conway Hart’s photography business began housing the business of a ‘Madame Nielly’, a French milliner:

 

The photographer duo Bourne & Shepherd also captured Calcutta’s Esplanade Row in 1865. Source: oldindianphotos.in

 

A comparison of the photography shop on the album photo and a photo taken in 1865.

This shows that from at least 1865, about a year after Conway Hart’s death, his business was closed entirely or was at least not operating from the same premises.

The ancestry website MyHeritage.com allows users to submit their own family lineages. Conway Hart’s presence on MyHeritage allowed us to find a relative of the painter who had taken an interest in their family history. 

One user had uploaded a photo of Conway Hart that can’t be found anywhere else online. Her name is Pamela Webster.

 

A view of Conway Hart’s profile on MyHeritage, managed by Pamela Webster.


Bellingcat reached out to Webster who  is the great-great-granddaughter of Conway Hart, and an artist herself.

She is working on a book about Conway for her grandchildren, and was happy to learn about the existence of the photo of his studio.

Back to the Archives

As museums continue to digitise their collections, artwork becomes more easily accessible for online researchers.

Connecting all of these dots is important as it helps to establish not just the location, period, and potential author of a historical photo, but other art as well. It can also help correct information that might have prevented others from making connections. 

For example, we can demonstrate that photos attributed to Conway Hart cannot have been taken in the 1870s, even though the Getty Museum in the US and London’s National Portrait Gallery have dated them as such.

The Metropolitan Museum in New York also appears to have a picture of Esplanade Row from the 1850s, again made by an unknown author. The picture mentions the building ‘Mountain Hotel’, though this is in fact the ‘Mountains Hotel’. It also doesn’t include the street name, only ‘Calcutta’.

But this skillset has another public interest application, too: a better understanding of how art can be traced online can help investigate lost and stolen art. The FBI’s website contains photos of artworks believed to be stolen, as does the German Lost Art Foundation which seeks to restitute artworks looted under Nazi rule

In some cases, a visit to the archives may still be necessary. Though online research cannot fully replace traditional research, the growth of online tools and new investigative methods can yield new clues about artworks and documents which have long mystified researchers and artists alike.

 


Bellingcat is a non-profit and the ability to carry out our work is dependent on the kind support of individual donors. If you would like to support our work, you can do so here. You can also subscribe to our Patreon channel here. Subscribe to our Newsletter and follow us on Twitter here.