Founded by a group of geoscientists, Australian startup Datarock is applying computer vision technology to the mining industry. More specifically, its deep learning models are helping geologists to analyze drill core sample images faster than before.
Typically, a geologist would examine these samples inch by inch to assess the mineralogy and structure, while engineers would look for physical characteristics such as flaws, fractures and rock quality. This process is slow and subject to human error.
"A computer can see rocks like an engineer would," Brenton Crawford, COO of Datarock, told InfoWorld . "If you can see it in the image, we can train a model to analyze it as well as a human."
Similar to Blue River, Datarock uses a variant of the RCNN model in production, with researchers turning to data augmentation techniques to gather enough training data in the early stages.
“After the initial discovery period, the team started combining techniques to create an image processing workflow for core drilling images. This involved the development of a series of deep learning models, which could process raw images in a structured format and segment important geological information, ”wrote the researchers in a post .
Using Datarock technology, customers can get results in half an hour, as opposed to the five or six hours it takes to record findings manually. This frees geologists from the most difficult parts, Crawford said. However, “when we automate things that are more difficult, we get some resistance and we have to explain that they are part of this system to train the models and make the feedback cycle turn”.
Like many companies that train deep learning computer vision models, Datarock started with TensorFlow, but soon moved to PyTorch.
"In the beginning, we used TensorFlow and it crashed with us for mysterious reasons," Duy Tin Truong, machine learning leader at Datarock, told InfoWorld . “PyTorch and Detecton2 were launched at that time and fit well with our needs. So, after some tests, we saw that it was easier to debug and work and took up less memory, so we converted, ”he said.
Datarock also reported a 4x improvement in the inference performance of TensorFlow for PyTorch and Detectron2 when running the models on GPUs - and 3x on CPUs.
Truong cited the growing PyTorch community, well-designed interface, ease of use and better debugging as reasons for switching and noted that, although “they are quite different from an interface point of view, if you know TensorFlow, switching is very easy, especially if you know Python . "
The IT Analyst will collect information from end users about system performance. They over the projects and handle the upgrades, hardware, and software installations.