摘要 |
Provided are an object detection method and device for online training. The method comprises: by acquiring an Nth frame image, and performing object detection and/or object tracking on the Nth frame image, acquiring at least one Nth object region; then, performing online training on a training detection unit by using an N-1th object region, wherein all the object regions detected in the first frame image to the N-1th frame image are stored in the training detection unit, thereby realizing parallel online processing of a detected object and the training detection unit, obtaining a plurality of object samples according to the needs of a user, meeting the needs of of user of an intelligent terminal for object diversification, so that accurate object regions can be obtained according to the current requirements of the user. |