О проекте

We develop an interactive semantic segmentation system which efficiently analyzes road pavement photos. First, a user gets a piece of an image and marks road marking and road defects on it, then the classifier is being learned. The next image part is pretended to be classified, so the user should correct classification errors. After this correction the classifier is learned again. While more image parts are being handled, number of mistakes decreases. See our paper for more details.

P Road monitoring


Нет активных обсуждений