Intelligent Annotation Algorithm Based on Deep-Sea Macrobenthic Images

keywords: Intelligent labeling, active learning, pseudo-labeling, object detection
In the field of image processing, due to the need of expertise and skills in deep-sea biology and the disadvantages of high labor cost and long time consuming, it has always been a difficult task to mark the images of deep-sea benthic organisms. To solve this problem, this paper proposes a new image intelligent labeling algorithm LACP AL (Localization-Aware-Choice and Pseudo Label Active Learning) which is based on Localization-Aware Active Learning. LACP AL is an active learning framework based on Faster R-CNN, it finds the ``valuable'' samples from unlabeled samples by clustering algorithm for every training; it selects hard-to-identify samples for manual annotation and further optimizes the model; and it proposes an improved pseudo-labeling mechanism to expand the training set and improve the model accuracy. According to the publicly available dataset provided by 2020 China Underwater Robot Professional Contest, a series of experiments has been done to verify that our algorithm can achieve higher recognition accuracy with fewer training samples compared with the existing algorithms for Marine benthic image recognition.
mathematics subject classification 2000: 68U10
reference: Vol. 41, 2022, No. 3, pp. 739–756