In the first 1.5 months, the device developed by Dmitry was fitted with a simple paper clip and attached to the horse’s leash. In this way the first (so far integral) datasets were obtained, on the basis of which during the first phase AI of the first phase Gilberto learned to distinguish the gaits of the horse with an accuracy of 89%.
And then a miracle happened! It was a real EVENT in the project.
Already after the first 25 sessions of training, Gilberto had a 95% probability of distinguishing one horse from another. From the horse’s signature …
The hypothesis of each horse’s individual gait was confirmed!
The final decision to work with Gilberto was made after 5 sets of data were provided for analysis. The task was to break down the sets and determine which data set, which horse corresponds to which one.
The diagnostic result was almost 98% and the project manager was convinced – that Gilberto is ideal for the team.
Now he is not just an employee, but an associate and member of our team.During this period, the project worked with a single sensor. However, for the tasks we had set ourselves, this data seemed insufficient.
First device prototype
The next stage of the project was the development of specialized sensors which could be attached to the horse’s leg and receive datasets from all 4 limbs. First version of the sensor was far from perfect… Rather big body, charging port of miniUSB, data recording to memory card – that’s what first sensor was.
But working with them, too, enabled a giant step forward in the development of the project. We suddenly discovered that the unique horse’s gait signature was already 100% accurate.
It was at this point that the project started, which involved analysing the trajectory of each limb and trying to understand what lies behind these seemingly random, but in fact perfectly defined and predictable movements…