Object detection models perform well at localizing and classifying objects that they are shown during training. However, due to the difficulty and cost associated with creating and annotating detection datasets, trained models detect a limited number of object types with unknown objects treated as background content. This hinders the adoption of conventional detectors in real-world applications like large-scale object matching, visual grounding, visual relation prediction, obstacle detection (where it is more important to determine the presence and location of objects than to find specific types), etc. We propose class-agnostic object detection as a new problem that focuses on detecting objects irrespective of their object-classes. Specifically, the goal is to predict bounding boxes for all objects in an image but not their object-classes. The predicted boxes can then be consumed by another system to perform application-specific classification, retrieval, etc. We propose training and evaluation protocols for benchmarking class-agnostic detectors to advance future research in this domain. Finally, we propose (1) baseline methods and (2) a new adversarial learning framework for class-agnostic detection that forces the model to exclude class-specific information from features used for predictions. Experimental results show that adversarial learning improves class-agnostic detection efficacy.
翻译:然而,由于建立和说明探测数据集的难度和成本,经过培训的模型检测了数量有限的物体类型,其背景内容被作为未知物体。这妨碍在现实世界应用中采用常规探测器,例如大型天体匹配、视觉地面定位、视觉关系预测、障碍探测(在确定天体的存在和位置比找到特定类型更重要的地方)等。我们提议,等级不可知物体探测是一个新问题,重点是探测物体,而不论其对象类别如何。具体而言,目标是预测图像中所有物体的捆绑框,而不是其目标类别。然后,预测的盒子会被另一个系统用于进行具体应用的分类、检索等。我们建议,为确定等级不可知探测器的基准制定培训和评价程序,以推进这一领域的未来研究。最后,我们提议(1) 基线方法和(2) 新的等级不可知性检测对抗性学习框架,迫使模型从预测中使用的特性中排除特定等级信息。实验性检测结果显示,对抗性测试将改进等级学习。