One of our best ever projects was working on the newly discovered Halma Collection – a vast collection of over 760 portraits of Ashburton locals found under a shed – they were captivated by the details of the people seen in the images.
It was as if past Ashburton residents had come to life, seeing their faces, expressions, and interests captured frame-by-frame on glass plate negatives.
Yet sometimes, seeing images like these also makes it easy to imagine that the past mostly happened in shades of grey, black and white. Of course, we know this isn’t true: as the variety of garments and colourful objects in the museum collection shows, colour was just as abundant in people’s lives in the past, as it is today.
While it’s not yet possible to bring back all the colourful details of the Halma Collection portraits, new digital technology is leading the way. Museum staff have been very proud to be part of an international project that is bringing Ashburton people back in splendid colour.
The project begins
The colourisation project has been undertaken by Andrew Jones, a London-based Kiwi working in the field of artificial intelligence.
In March this year, the museum team read an article by Aaron Goile on the Stuff news website about how Jones had colourised images of historic All Blacks. The results were spectacular – bringing these famous sportsmen back to life in colour to show them as we had never imagined them to look before. For the first time, people could see historic players such as George Nepia, Dave Gallaher and Cyril Brownlie in more detail than ever before.
Museum staff contacted Jones, hoping he might do the something similar for a few images from the Halma Collection. It was big ask – and very exciting when he said ‘yes’.
Originally from Matamata, Jones is a graduate of Victoria University in Wellington. He has been working on analytics and data science projects for the past decade, including for companies like Amazon and Sony PlayStation. We are very fortunate to be working with him and helping our community gain from his expertise.
Over the past few weeks Jones and museum staff have been sending images in batches across the world to bring back details of this extraordinary collection.
These images have come from the scans of glass plate negatives, converted to positive digital images. Just a few of the results can be seen on this page.
To colour black and white images, Jones has developed a programme using artificial intelligence. He uses an algorithm to colourise the photos that have been created from scans of the Halma Collection negatives.
To colour each image pixel, Jones must programme his computer to recognise up to 16.7 million colour options, which in the black and white images appear only as shades of grey.
Every pixel is made up of a mix of red, green or blue. These three colours each represent a value from zero to 255, and the combination of the three values determines the final colour that the programme will apply to that pixel.
It is because there are so many options across the three colour combinations that the over 16 million colour variations are the result.
The programme is also taught to recognise features such as faces, plants and so forth to improve the colour recognition. This has resulted in the programme being able to distinguish that the details of plants but not faces are shades of green, with remarkable accuracy.
Jones programme is based on an artificial intelligence model, called a neural network. It’s a mathematical model built to try and mimic how our human brains learn.
It’s an ever evolving programme and project for Jones, to turn abstract data that a computer understands into the familiar faces of people who have been loved, lived and rediscovered in Ashburton.
By Tanya Zoe Robinson
- Images from the Halma Collection alongside the colourised version created through the programme devised by London-based Kiwi, Andrew Jones.