Interpolation-based semi-supervised learning for object detection
This work addresses the problem of reducing labeling costs for object detection tasks, which is incremental as it adapts existing interpolation techniques to a specific domain.
They tackled the high labeling cost in object detection by proposing an interpolation-based semi-supervised learning method (ISD) that addresses issues with applying conventional interpolation regularization directly, resulting in significant performance improvements on benchmark datasets like PASCAL VOC and MSCOCO.
Despite the data labeling cost for the object detection tasks being substantially more than that of the classification tasks, semi-supervised learning methods for object detection have not been studied much. In this paper, we propose an Interpolation-based Semi-supervised learning method for object Detection (ISD), which considers and solves the problems caused by applying conventional Interpolation Regularization (IR) directly to object detection. We divide the output of the model into two types according to the objectness scores of both original patches that are mixed in IR. Then, we apply a separate loss suitable for each type in an unsupervised manner. The proposed losses dramatically improve the performance of semi-supervised learning as well as supervised learning. In the supervised learning setting, our method improves the baseline methods by a significant margin. In the semi-supervised learning setting, our algorithm improves the performance on a benchmark dataset (PASCAL VOC and MSCOCO) in a benchmark architecture (SSD).