Bring new insights to image data
Computer vision and labelling bring new insights to image data. Incorporate image data into your workflows to expand your understanding.
EarthVISION enables labelling and ML interpretation of image data. With generic applicability across images of different types, it includes functionality for image classification, object detection, unsupervised image segmentation, semantic embedding, semi-supervised segmentation and supervised segmentation.
Initial development is centred around interpretation of rock cuttings images in collaboration with Rockwash Geodata and Lundin. Available to a consortium of companies for interpretation of all drill cuttings data in Norway (over 700 000 samples), as well as companies worldwide looking to exploit the highly valuable, but traditionally dormant, data from cuttings samples.
Ongoing development will address applications for other types of image data e.g.
interpretation of additional rock images from microscopy and core
interpretation of seabed data such as bathymetry and side-scan sonar
interpretation of submarine imagery from ROVs and AUVs
During image classification of drill cuttings samples, the labelling and deep-learning prediction steps are supported by the combination of image data with a rich suite of metadata including, stratigraphic information, elemental composition (XRF), mineralogical composition (XRD) and more.
Image segmentation is a means to extract very detailed information from images. This workflow includes rapid annotation for image segmentation as well as unsupervised segmentation, semantic embedding and training of supervised segmentation models.
Image segmentation enables localisation and classification of minerals and lithologies with pixel-level accuracy. Percentages of minerals and lithologies can be computed from the segmentation mask and the data can be transferred to the well-log domain – thus enabling integration of image data with the well-log and seismic data.