A recommended object determination method includes: acquiring at least one key frame corresponding to video data, the key frame being a video frame that validly identifies an in-frame object; performing an object detection operation on a first key frame among the key frames to determine object information associated with the first key frame, the object information including: object semantic information and object visual information; and determining candidate objects based on the object semantic information, and obtaining a recommended object associated with the first key frame by screening the candidate objects based on the object visual information. A user may be provided with a object precisely associated with video content, thus improving user experience.