How AI is creating an impact in opal mining – ARNnet

Launching a new tool OpalAI, Dan Brown has managed to combine his two passions. 

When he’s not putting his Microsoft expertise to use, Adelaide-based MVP Daniel Brown has another passionate project – opal mining. 

With his launch of a new tool OpalAI, he has managed to combine his two passions. 

Twice per year, Brown’s other venture drills for opal and developed the tool to detect opal using image recognition on drilled material. 

The idea was to analyse rocks ejected by a spinoff drill to detect opal visually, flagging any unknown or opal-bearing rocks, and automatically alerting team members to reduce the downtime of manually searching through piles.

OpalAI is built on Azure AI’s Computer Vision and Machine Learning services. Using a database of over 3000 images of opal in various states—within rock walls, loose, polished, and set in jewellery—the model can detect opal in diverse situations with high accuracy.

Additionally, OpalAI has been trained to leverage the unique way opal fluoresces under UV light, enhancing detection in low-light settings.

OpalAI’s real-time visual detection capabilities reduce the risk of missing valuable material. 

“This added layer of detection could make the mining process more efficient and reliable,” Brown said. “My experience with Microsoft’s Azure AI has been instrumental in bringing OpalAI to life, from training the model to identifying opal with precision in challenging environments.”

One of the biggest challenges faced was obtaining photos of opal embedded in rock as miners are reluctant to reveal opal in its natural state. 

“Through persistence, we built relationships with a few miners who supplied valuable images and videos,” he said.

“Environmental factors, like intense sunlight on screens, dust, and flies, also posed issues. A simple fix – using a blanket over the laptop – with visibility, while fly nets and frequent cleaning became essential practices.”

OpalAI’s capabilities as an “extra set of eyes” are valuable in mining, Brown said and its model is currently also trained to detect humans, machinery, and hazards like trenches.

“With further development, it can support hazard identification and safety procedures information presentation on-screen,” he said. 

Brown and OpalAI will also appear in an episode for Outback Opal Hunters which is currently airing in the UK, with the series set to be on TV sometime in 2025 in Australia. 

“During the filming of Opal Hunters in October 2023, I was down a shaft with the laptop and we had a successful detection of Opal which was captured on the cameras,” he said. “When we returned to Coober Pedy, we had our second successful detection of Opal – colour – which was amazing.”

Brown said he envisions OpalAI evolving to tackle other challenges in mining by refining its visual detection capabilities, enabling it to identify a wider range of minerals beyond opal. 

“As its image recognition improves, OpalAI could assist miners in rapidly assessing excavation sites for valuable deposits, streamlining the process and reducing the time spent on exploratory digging,” he said. 

“The model itself can be used to detect a wide variety and range of gemstones, provided there is enough training material of high quality in its reference database.”