In this context, our company Geo4i has been working for a few months now in this field, in order to improve its geospatial analysis capabilities.

Even if the human remains at the heart of image intelligence and if the image exists to answer a (precise) question posed by a customer.

As part of the development of our automatic processing workflow (our aim is to accelerate the imagery processing part), It seemed interesting to us to invest on this technique in order to continue to refocus the analyst on its core business: ANALYSIS.

Thus, after becoming familiar with the subject, we have tried various approaches to obtain the best results in terms of equipment identification, as the continuity of our identification tool, Help4i.

The process is currently integrated into our platform, GEOSPACE, in a beta version and the success rate is around 85%.

Even if the results obtained are still insufficient for us, the recently acquired experience in this technology allowed us to redefine a new method of deep learning processing. That should allow us to significantly increase the rate of equipment identification.

Of course, the work has already began and is rather conclusive. 😊