Method and apparatus with traffic light recognition model

An electronic device comprising:

• a camera;
• one or more processors;
• and
• a memory storing instructions configured to cause the one or more processors to:
• obtain an input image from the camera;
• estimate, based on prior information about traffic light objects, a first region of interest (RoI) for the input image;
• determine a second RoI based on the first RoI and based on detection results of previous frames of the camera, wherein the detection results of the previous frames correspond to recognitions results of recognizing traffic lights in the previous frames by an object recognition model configured to recognize traffic lights;
• recognize, by the object recognition model, a traffic light in the input image, wherein the recognizing is based on the input image and the second RoI;
• and
• control an operation of a vehicle, based on the result of the recognition,
• wherein, for determining the second RoI, the instructions are further configured to cause the one or more processors to aggregate the detection results of the previous frames, obtain representative centers of the detection results of the previous frames, filter the centers, obtain the second RoI based on the filtered centers, and determine the second RoI based thereon.

A method performed by a computing device, the method comprising:

• receiving input images captured by a camera of a moving vehicle;
• for a first of the input images, determining a first region of interest (RoI), wherein the first RoI is determined based on historical traffic light observations;
• inputting, to an object recognition model configured to recognize traffic lights, the first input image and an enlargement of a region in the first input image defined by the first RoI, based on which the object recognition model infers a location of a first traffic light in the first input image;
• for a second of the input images captured after the first input image, determining a second RoI, wherein the second RoI is determined based on the location of the traffic light in the first input image, and wherein the determining the second RoI comprises aggregating the detection results of the previous frames, obtaining representative centers of the detection results of the previous frames, filtering the centers, obtaining the second RoI based on the filtered centers, and determining the second RoI based on the centers of clusters of the aggregated detection results;
• inputting, to the object recognition model, the second input image and an enlargement of a region in the second input image defined by the second RoI, based on which the object recognition model infers a location, in the second input image, of the first traffic light or a second traffic light;
• and
• controlling an operation of the moving vehicle based on the inferred location in the second input image.