A person’s signature is a representative of his identity. For us at the Bank, a signed document by a customer is an instruction from him for carrying out an approved transaction for him.

On on-boarding a customer we capture an image of his signature in our systems, and on receiving a signed document (Cheques, DDs and others) from him we match the signature on the document with the one recorded in the database before proceeding with the instruction.

In the case of skilled forgeries, it becomes very difficult to verify the identity of the customer.

We want you to build a system that can help us distinguish forgeries from actual signatures. This system should be able to study signature parameters as strokes, curves, dots, dashes, writing fluidity & style, in a Writer-Independent manner and create features for identification of the signature.

The system should not use any existing APIs and should be completely self-developed.

How should it work?

The system shall work in 2 steps:

• Step 1: Accept & Store Genuine Signature Image: Take actual signature scanned image of the on-boarding customer and store it in a database against a unique Customer ID
• Step 2: Accept & Compare Signature Images: Accept inputs of Customer ID, and corresponding signature image. Compare with the signature stored in DB against the given Customer ID, and return a Confidence Match Score between the two signature images

Download Sample Datasets

With advent of AI-RPA, document reading and analysis are on their way for becoming mainstream. Tables in documents help represent information in a structured way. For a comprehensive document analysis, it is important to be able to identify, read and understand tables present in documents. This step, although despite a successful OCR, may become difficult, as tables vary a lot in their layout, formats and encoding.

The problem at hand is to build a solution that can detect tables in given documents, then make sense of the information they present.

For this problem too, you cannot use any existing APIs.

How should the system work?

• Step 1: Accept document input, read tables: System should have an input mechanism for accepting documents (PDFs, MSWord) or document images (TIFF, JPEG). The document may have one or more tables.
• Step 2: Step 2: As an output, system should return the table content in an excel format,same as that in the sample data sets
  Download Sample Datasets