Aktif Öǧrenme Yöntemi Kullanarak Nesne Tespiti

In the last decade, deep learning-based object detection models have achieved high performance. However, to train these object detection models, a large amount of labeled images is required. Active learning is a machine learning procedure that is useful in reducing the amount of labeled data required to achieve the targeted performance. With active learning, it is possible to obtain high performing models on real-world data where annotation is time-consuming, while decreasing the labeling cost. It helps reduce the cost of data labeling by efficiently selecting a subset of informative samples from a large repository of unlabeled data. In this study, we developed an object detection model combined with active learning. The results of the experiments show that almost the same level of success was achieved by labeling a smaller amount of data with the active learning framework, compared to labeling and using all the data, leading to lower labeling costs.