Method, system and apparatus for image orientation correction

A client device for generating a training dataset for training a machine learning, ML, model using federated learning to perform image orientation correction, the client device comprising:

• an image capture device, for capturing an image of a scene;
• a sensor for providing a predicted tilt angle of the scene in the captured image;
• storage comprising a database forming a training dataset for training the ML model;
• and
• at least one processor coupled to memory and arranged to:
• determine whether the captured image and predicted tilt angle improves a non-bias score of the database, and
• store metadata associated with the captured image in the database when the captured image is determined to improve the non-bias score of the database,
• wherein the non-bias score for the database is obtained based on a difference in dispersion between the database and an unbiased dataset.

A computer-implemented method, performed by a client device, for training a machine learning, ML, model using federated learning to perform image orientation correction, the method comprising:

• obtaining a ML model for training locally by the client device;
• checking whether computational resources of the client device are available to train the ML model;
• and
• training the ML model, when the resources are available, using a training dataset on the client device, the training dataset comprising metadata associated with images having a range of tilt angles,
• wherein the ML model is trained based on a non-bias score for the database obtained based on a difference in dispersion between the training dataset and an unbiased dataset.