Method and system for creating training data

A method for creating training data for a machine learning algorithm, comprising: displaying 204, for a user, an image 202 containing an elongate object 206 defining a path between a first end 208 and a second end 210; and creating data points (nodes) 216 and labels 221 along the object's path via a user interaction component 220. The first data point may be placed within a second threshold distance of the first end of the object. The location of each node and line may be between the object's edges 207. Each node (data point) and label may be created by a single press of a button on the user interaction component. Each label may indicate the visibility of a portion of the object as defined by the line linking nodes. A first button press may place a label indicating the corresponding object portion is visible and a second button press may indicate an occluded or not visible segment. A third button press may place a final node within the first threshold distance of the second end of the object. The method may be applied to create training images for a classifier that detects seat belts in a vehicle.