One-Shot learning recognition

Realization of own original recognition algorithms. Unlike neural networks and depp learning, you do not need many and many objects for training. It is enough for one sample and one action for the object to be saved and to be recognized

Features:

• Recognizes any objects by characteristic outlines in photos (restrictions below)

• It does not depend on the position and size of the object. The algorithm is oriented only on the proportions

• It is enough to remember the object once to find a similar object

• Based on this algorithm, an API can be proposed for embedding in its solutions

Training and recognition. Restrictions.

Objects for training should be in the center of the frame and fill most of the frame. Preferably on a neutral background

In the current version, the pictures are converted to a size of 100 * 100 pixels so complex objects can not be recognized