Method and apparatus for detecting mobile traffic light

An apparatus for detecting a mobile traffic light, comprising:

• at least one processor;
• and
• a memory communicatively coupled to the at least one processor,
• wherein the at least one processor is configured to:
• obtain a set of positive sample images with the mobile traffic light and a set of negative sample images without the mobile traffic light;
• extract current training data respectively from the set of positive sample images and the set of negative sample images based on a preset sampling ratio;
• perform detection on the mobile traffic light of each image in the current training data with a first preset detection model to determine a detection result of each image;
• determine current classification loss and current detecting frame regression loss based on a difference between the detection result of each image and labeled data;
• and
• generating a detection model for detecting the mobile traffic light in case that both the current classification loss and the current detecting frame regression loss meet a preset condition.

A non-transitory computer readable storage medium having computer instructions stored thereon, wherein the computer instructions are configured to cause a computer to execute a method for detecting a mobile traffic light, the method comprising:

• obtaining a set of positive sample images with the mobile traffic light and a set of negative sample images without the mobile traffic light;
• extracting current training data respectively from the set of positive sample images and the set of negative sample images based on a preset sampling ratio;
• and
• training with a first preset detection model to generate a detection model, which comprises:
• performing detection on the mobile traffic light of each image in the current training data with the first preset detection model to determine a first detection result of each image;
• determining current classification loss and current detecting frame regression loss based on a difference between the first detection result of each image and labeled data;
• and
• generating the detection model for detecting the mobile traffic light in case that both the current classification loss and the current detecting frame regression loss meet a preset condition.