Use case
Aerial images taken by drone, satellite photos, ground cameras, provide an excellent way to monitor the condition of crops or the extension of vegetation. This visual data can further be exploited to make decisions.
Instead of going through the images one by one, it is possible to train an artificial neural network to recognize and classify the elements it detects on the images. The technologies are identical to those that Logiroad.ai has been deploying for several years to monitor the condition of roads or vegetation along their edges.
This work requires annotating a number of images, to give the neural network examples to learn from. Then, an A.I. adapted to the problem is developed.
It is possible to “look inside” the neural network to understand how decisions are made. For example, we can identify which regions of the image led to which decision. Another example, we can estimate the uncertainty of the output to know the robustness of the predictions.