Object detection is widely studied in computer vision filed. In recent years, certain representative deep learning based detection methods along with solid benchmarks are proposed, which boosts the development of related researchs. However, there is no object detection benchmark targeted at military field so far. To facilitate future military object detection research, we propose a novel, publicly available object detection benchmark in military filed called MOD, which contains 6,000 images and 17,465 labeled instances. Unlike previous benchmarks, objects in MOD contain unique challenges such as camouflage, blur, inter-class similarity, intra-class variance and complex military environment. Experiments show that under above chanllenges, existing detection methods suffer from undesirable performance. To address this issue, we propose LGA-RCNN which utilizes a loss-guided attention (LGA) module to highlight representative region of objects. Then, those highlighted local information are fused with global information for precise classification and localization. Extensive experiments on MOD validate the effectiveness of our method.
翻译:近些年来,提出了某些有代表性的深层次学习型探测方法和坚实的基准,这促进了相关研究的发展。然而,迄今为止,还没有针对军事领域的物体探测基准。为了便利未来的军事物体探测研究,我们提议在军事档案中建立一个新的、公开的、称为MOD的物体探测基准,其中包括6,000个图像和17,465个标记实例。与以往的基准不同,MOD的物体含有独特的挑战,如迷彩、模糊、阶级间相似、阶级间差异和复杂的军事环境。实验表明,在上文的Chanllenge中,现有的探测方法存在不良性能。为了解决这一问题,我们建议LGA-RCNNN采用一个损失引导注意模块来突出具有代表性的物体区域。然后,那些突出的地方信息与全球信息相结合,以便精确分类和本地化。关于MOD的广泛实验证实了我们的方法的有效性。