We present the first systematic study on concealed object detection (COD),which aims to identify objects that are "perfectly" embedded in theirbackground. The high intrinsic similarities between the concealed objects andtheir background make COD far more challenging than traditional objectdetection/segmentation. To better understand this task, we collect alarge-scale dataset, called COD10K, which consists of 10,000 images coveringconcealed objects in diverse real-world scenarios from 78 object categories.Further, we provide rich annotations including object categories, objectboundaries, challenging attributes, object-level labels, and instance-levelannotations. Our COD10K is the largest COD dataset to date, with the richestannotations, which enables comprehensive concealed object understanding and caneven be used to help progress several other vision tasks, such as detection,segmentation, classification, etc. Motivated by how animals hunt in the wild,we also design a simple but strong baseline for COD, termed the SearchIdentification Network (SINet). Without any bells and whistles, SINetoutperforms 12 cutting-edge baselines on all datasets tested, making themrobust, general architectures that could serve as catalysts for future researchin COD. Finally, we provide some interesting findings and highlight severalpotential applications and future directions. To spark research in this newfield, our code, dataset, and online demo are available on our project page:http://mmcheng.net/cod.