Joint evaluation method for dual-network model, equipment and storage medium

Abstract

The invention relates to the field of camera detection, and discloses a joint evaluation method for a dual-network model, equipment and a storage medium. The method comprises the following steps: acquiring a plurality of images shot by a vehicle-mounted BSD camera; identifying each image by adopting a semantic segmentation model and a target detection model, and determining whether effective targets with the same category and corresponding positions exist or not; scoring each image according to whether the effective target exists in each image or not and whether the effective target is matched with the alarm target marked in the alarm area of the same image or not under the condition that the effective target exists in each image; and summarizing scores of the multiple images, and determining evaluation results of the semantic segmentation model and the target detection model according to the scores. According to the embodiment, joint evaluation of the semantic segmentation model and the target detection model is realized.

Description

technical field [0001] The invention relates to the field of camera detection, in particular to a joint evaluation method, device and storage medium of a dual network model. Background technique [0002] The vehicle-mounted BSD (Blind Spot Detection, blind spot monitoring) camera (hereinafter referred to as the camera) is installed on both sides of the rear of the vehicle, and is used to detect the blind spots on both sides of the rear of the vehicle when the vehicle is driving. When the target object, such as other vehicles or pedestrians, is detected in the image captured by the camera, an alarm prompt will be given. [0003] At present, image processing is generally performed through semantic segmentation models and object detection models to identify target objects. The accuracy of these two models will significantly affect the accuracy of recognition, resulting in false or missed alarms. The conventional model evaluation method is to evaluate the precision and recall ...

AI Technical Summary

Problems solved by technology

Among them, the labor cost and time cost of the road test method are high, the reliability is not high, and mistakes are prone to occur, resulting in deviations in the final evaluation results.

The way the bounding box is evaluated is only suitable for single model evaluation and cannot measure the results of dual models, and the evaluation of a single model does not provide intuitive information about the correct / wrong object of the detected object, when the two models show different accuracy for different objects When , even if the overall accuracy of any model is high, the accuracy of its combination is not ideal; the evaluation flexibility of a single model is poor, and they cannot accumulate and count the test information of the region of interest on multiple images

In addition, quantitative and qualitative assessments are often mixed, resulting in ambiguity in measurement

Images