An Automatic Annotation Algorithm for Deep Learning Image Datasets Based on HOG Features

Abstract

Labeled data is crucial for deep learning model training, but it is mainly based on manual annotation. This paper proposes an automatic annotation algorithm for deep learning image datasets based on HOG features. The algorithm obtains the position information of the target based on the Gaussian mixture model, and automatically obtains the target category based on the Mean Shift algorithm of HOG feature. The labeled sample obtained in the above steps is used to train the Deformable Part Model to annotate more unlabeled image data. Experimental results show that this algorithm can quickly complete the automatic annotation of image datasets, reduce the cost of acquiring labeled datasets and greatly improve the effectiveness of acquiring labeled datasets in deep learning.