Dynamic signature identification from handwritten elements

A system comprising:

• one or more processors;
• and
• a memory coupled to the one or more processors, the memory storing a plurality of instructions executable by the one or more processors, the plurality of instructions comprising instructions executable by the one or more processors to cause the system to perform operations comprising:
• identifying a set of handwritten elements from a document image;
• generating, using a trained model, a set of embeddings for the set of handwritten elements by generating an embedding for each handwritten element in the set of handwritten elements, wherein an embedding generated for a handwritten element represents one or more features extracted from the handwritten element;
• clustering the set of embeddings to create a set of clusters, each cluster included in the set of clusters comprising at least one embedding from the set of embeddings;
• processing the document image to identify one or more areas within the document image designated for placement of a signature;
• based upon locations of the one or more areas within the document image, tagging each cluster included in the set of clusters as a signature cluster or as a non-signature cluster;
• for each cluster included in the set of clusters that is tagged as a signature cluster, and for each embedding included in the signature cluster, outputting an indication that a handwritten element of the set of handwritten elements corresponding to the embedding is a signature element;
• and
• for each cluster included in the set of clusters that is tagged as a non-signature cluster, and for each embedding included in the non-signature cluster, outputting an indication that a handwritten element of the set of handwritten elements corresponding to the embedding is a non-signature element.

A non-transitory computer-readable memory storing a plurality of instructions executable by one or more processors, the plurality of instructions comprising instructions that, when executed by the one or more processors, cause performance of operations comprising:

• identifying a set of handwritten elements from a document image;
• generating, using a trained model, a set of embeddings for the set of handwritten elements by generating an embedding for each handwritten element in the set of handwritten elements, wherein an embedding generated for a handwritten element represents one or more features extracted from the handwritten element;
• clustering the set of embeddings to create a set of clusters, each cluster included in the set of clusters comprising at least one embedding from the set of embeddings;
• processing the document image to identify one or more areas within the document image designated for placement of a signature;
• based upon locations of the one or more areas within the document image, tagging each cluster included in the set of clusters as a signature cluster or as a non-signature cluster;
• for each cluster included in the set of clusters that is tagged as a signature cluster, and for each embedding included in the signature cluster, outputting an indication that a handwritten element of the set of handwritten elements corresponding to the embedding is a signature element;
• and
• for each cluster included in the set of clusters that is tagged as a non-signature cluster, and for each embedding included in the non-signature cluster, outputting an indication that a handwritten element of the set of handwritten elements corresponding to the embedding is a non-signature element.