The Elephant in the Room
This reveals vulnerabilities in object detectors, which is important for computer vision researchers and practitioners, though it is incremental as it focuses on analyzing existing failures rather than proposing a new solution.
The paper identifies a family of failures in state-of-the-art object detectors caused by 'object transplanting', where replacing image sub-regions with other trained objects leads to non-local impacts, such as changes in object identity and detection of other objects.
We showcase a family of common failures of state-of-the art object detectors. These are obtained by replacing image sub-regions by another sub-image that contains a trained object. We call this "object transplanting". Modifying an image in this manner is shown to have a non-local impact on object detection. Slight changes in object position can affect its identity according to an object detector as well as that of other objects in the image. We provide some analysis and suggest possible reasons for the reported phenomena.