Cheque Signature Extraction

Introduction

The Cheque Signature Extraction API detects and extracts signature regions from cheque images. Using a computer-vision detection model, the API locates each signature on the cheque, returns its bounding box and provides a cropped image of the detected region encoded in base64.

This API helps organizations automate cheque processing, isolate signatures for downstream signature-verification workflows, and reduce manual review during clearing and onboarding.

Accuracy for signature detection - 95%

API Details

Endpoint:

 POST  https://api.signzy.app/api/v3/extraction/cheque-signature

POST  https://api-preproduction.signzy.app/api/v3/extraction/cheque-signature

Request Parameters

Request Body Parameters

Parameter	Type	Required	Description
input	object	Yes	Wrapper object containing the request parameters.
input.image	string	Yes	Cheque image to process. Supported file types: jpg, jpeg, png. (Persisted file URL)
input.confidence_threshold	number	Yes	Minimum confidence a detection must meet to be returned. Value between 0 and 1. Example: 0.6 .

Request Headers

Name	Value	Required	Description
Content-Type	application/json	Yes	The type of content that the request body contains.
Authorization	XXXXXXXXXXX	Yes	An authentication token to authorize the request. Reach out to the Signzy support team to get one created.

JSON

{ "input": { "confidence_threshold": 0.6, "image": "image url" } }

Code samples

curl --location 'https://api-preproduction.signzy.app/api/v3/extraction/cheque-signature' \
--header 'Content-Type: application/json' \
--header 'Authorization: <auth_key>' \
--data '{
    "input": {
        "confidence_threshold": 0.6,
        "image": "image url"
    }
}'

Response body parameters

Group	Parameter	Data Type	Description
execution	execution_id	string	Unique identifier for this execution. Example: 2fd0f340-2f86-40bf-adc4-0f8e7e6156bc.
execution	status	string	Processing status of the request. Example: success.
output	output.detection_count	integer	Number of signature regions detected that met the confidence threshold.
detections	output.detections.label	string	Class label of the detected region. Example: Signature.
detections	output.detections.confidence	number	Model confidence for the detection, as a value between 0 and 1. Example: 0.85617.
detections	output.detections.bounding_box	object	Pixel coordinates of the detected region within the source image.
detections	output.detections.bounding_box.x	number	X coordinate of the top-left corner of the bounding box, in pixels.
detections	output.detections.bounding_box.y	number	Y coordinate of the top-left corner of the bounding box, in pixels.
detections	output.detections.bounding_box.width	number	Width of the bounding box, in pixels.
detections	output.detections.bounding_box.height	number	Height of the bounding box, in pixels.
detections	output.detections.cropped_image	string	Base64-encoded JPEG of the cropped signature region.

Response examples

//When detection count more than 0 
{
    "data": {
        "execution_id": "2fd0f340-2f86-40bf-adc4-0f8e7e6156bc",
        "status": "success",
        "output": {
            "detection_count": 1,
            "detections": {
                "label": "Signature",
                "confidence": 0.85617,
                "bounding_box": {
                    "x": 343.32581,
                    "y": 149.21405,
                    "width": 52.39206,
                    "height": 91.34175
                },
                "cropped_image": "/9j/4AAQSkZJRgABAgAAAQABAAD/wAARCABbADQDAREAAhEBAxEB..."
            }
        }
    }
}

//When detection count 0

{
    "data": {
        "execution_id": "b7438435-0eb0-4922-a90a-837e9275401c",
        "status": "success",
        "output": {
            "detection_count": 0
        }
    }
}

Best Practices

File Quality: Ensure cheques are clear, legible, and not heavily watermarked or damaged for optimal extraction accuracy.

Supported Formats: Use supported file formats (JPG, JPEG, PNG, PDF, TIFF) for best results

Contact Us for Any Assistance

If you have any questions or need assistance, please reach out to our customer support team. You can contact us via email at [email protected]. We strive to provide prompt and reliable assistance, ensuring your queries are addressed effectively.