Abstract: Various methods, apparatuses, and media for implementing a check fraud detection module are provided. A processor parses received digital image of a check into separate portions, one of the portions including a signature of an account holder. The processor applies a machine learning model to generate a new 128-dimensional embedding of the signature of the account holder parsed from the received digital image of the check and compares it preauthorized historical reference 128-dimensional embedding of the signature stored onto a database. The processor generates, based on comparing, a similarity score between the new 128-dimensional embedding of the signature and the preauthorized historical reference 128-dimensional embedding of the signature; and identifies whether the received check is fraudulent or not based on the generated similarity score.

Abstract: A method of selectively transferring micro devices from a donor substrate to contact pads on a receiver substrate. Micro devices being attached to a donor substrate with a donor force. The donor substrate and receiver substrate are aligned and brought together so that selected micro devices meet corresponding contact pads. A receiver force is generated to hold selected micro devices to the contact pads on the receiver substrate. The donor force is weakened and the substrates are moved apart leaving selected micro devices on the receiver substrate. Several methods of generating the receiver force are disclosed, including adhesive, mechanical and electrostatic techniques.