Hi, I've tried the VOD task before, but I stopped the study due to other higher priority projects. In short, I think the variant of StreamYOLO can deal with VOD task. The recent detectors of VOD are two-stage ones, if making the single-stage detectors to work in this field may be a good job. In addition, the VOD annotation information format is the same as VOC, which YOLOX has off-the-shelf interface to train.
Hi, I've tried the VOD task before, but I stopped the study due to other higher priority projects. In short, I think the variant of StreamYOLO can deal with VOD task. The recent detectors of VOD are two-stage ones, if making the single-stage detectors to work in this field may be a good job. In addition, the VOD annotation information format is the same as VOC, which YOLOX has off-the-shelf interface to train.