Method and device for image processing, method for training object detection model

A method for training an object detection model, comprising:

• acquiring a plurality of sets of training data, each set of training data comprising a visible light image, a depth image and a corresponding labeled object mask pattern for a same scenario;
• generating a central weight map corresponding to the visible light image, weight values represented by the central weight map gradually decreasing from a center to an edge of the visible light image;
• and
• performing training by taking the visible light image, the depth image and the central weight map as an input of an object detection model comprising an initial network weight and taking the labeled object mask pattern as a ground truth outputted by the object detection model comprising the initial network weight, to obtain a target network weight of the object detection model.

A device for image processing, comprising:

• a processor;
• and
• a memory having stored thereon a computer program,
• the computer program, when executed by the processor, enables the processor to implement operations of:
• acquiring a visible light image;
• generating a central weight map corresponding to the visible light image, weight values represented by the central weight map gradually decreasing from a center to an edge of the visible light image;
• inputting the visible light image and the central weight map into an object detection model to obtain an object region confidence map, wherein the object detection model is a model obtained by training according to a plurality of sets of training data, and each set of training data comprises a visible light image, a central weight map and a labeled object mask pattern for a same scenario;
• and
• determining at least one target object in the visible light image according to the object region confidence map.