# US 11,113,836 B2 · Object detection method, device, apparatus and computer-readable storage medium > Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt - Number: US 11,113,836 B2 - Document ID: US11113836B2 - Status: Granted - Kind: B2 - Application number: 16/512,249 - Filed: 2019-07-15 - Published: 2020-03-05 - Granted: 2021-09-07 - Earliest claimed: 2018-09-03 - Assignees: Baidu Online Network Technology (Beijing) Co., Ltd. - Applicants: Baidu Online Network Technology (Beijing) Co., Ltd. - Inventors: Xuehui Wang; Ming Li; Tian Xia - Web page: https://patentgrep.com/patent/US11113836B2 ## CPC - G06V 10/25: Determination of region of interest [ROI] or a volume of interest [VOI] - G06T 7/62: of area, perimeter, diameter or volume - G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting - G06V 20/10: Terrestrial scenes (scenes under surveillance with static cameras G06V20/52; scenes perceived from the exterior of a vehicle G06V20/56; scenes perceived from the interior of a vehicle G06V20/59) - G06T 2207/20081: Training; Learning - G06V 2201/07: Target detection - G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING - G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL ## Abstract Embodiments of object detection method, device, apparatus and a computer-readable storage medium are provided. The method can include: obtaining an enclosing frame of a target object in an input image; according to the enclosing frame, determining a reference frame from a predetermined candidate frame set comprising a plurality of candidate frames; generating a size-related feature according to a size of the reference frame and a size of the enclosing frame; and detecting an object in the input image by applying the size-related feature in a machine learning model. In an embodiment of the present application, the object detection is performed by using a feature related to an object size, that is, the prediction criterion related to the object size is added to an original feature in a machine learning model, thereby further improving the accuracy of the object detection. ## Claims ### Claim 1 An object detection method, comprising: obtaining an enclosing frame of a target object in an input image; according to the enclosing frame, calculating respective overlap degrees of the enclosing frame of the target object with respective candidate frames in a candidate frame set, the candidate frame set is a predetermined frame set associated with the category of the target object in the input image and each target object in the input image has one corresponding candidate frame set; determining a predetermined number of candidate frames with overlap degrees thereof ranked in the top as reference frames; generating a size-related feature according to respective sizes of the reference frames and a size of the enclosing frame; and detecting an object in the input image by applying the size-related feature in a machine learning model. ### Claim 2 (depends on 1) The object detection method according to claim 1, wherein each of the enclosing frame and the candidate frames is of a polygon shape, and the calculating respective overlap degrees of the enclosing frame of the target object with respective candidate frames in the candidate frame set comprises: for each candidate frame, aligning the enclosing frame with the candidate frame at one vertex; calculating an intersection-over-union of the enclosing frame with the candidate frame; and determining the calculated intersection-over-union as the overlap degree of the enclosing frame with the candidate frame. ### Claim 3 (depends on 1) The object detection method according to claim 1, wherein the determining a predetermined number of candidate frames with overlap degrees thereof ranked in the top as reference frames comprises: in a case that the overlap degree of the enclosing frame with the candidate frame is greater than an overlap threshold, determining the candidate frame as the reference frame. ### Claim 4 (depends on 1) The object detection method according to claim 1, wherein the generating a size-related feature according to respective sizes of the reference frames and a size of the enclosing frame comprises: determining a width ratio and a height ratio of the enclosing frame with the reference frame as the size-related feature, or determining a natural logarithm of the width ratio and a natural logarithm of the height ratio as the size-related feature. ### Claim 5 (depends on 1) The object detection method according to claim 1, wherein the method further comprises: predetermining a relationship between respective candidate frame sets and respective object categories; determining an object category to which the object to be detected belongs; and selecting a candidate frame set according to the determined object category. ### Claim 6 An object detection device, comprising: one or more processors; and a storage device configured for storing one or more programs, wherein the one or more programs, when executed by the one or more processors, cause the one or more processors to: obtain an enclosing frame of a target object in an input image; according to the enclosing frame, calculate respective overlap degrees of the enclosing frame of the target object with respective candidate frames in a candidate frame set, the candidate frame set is a predetermined frame set associated with the category of the target object in the input image and each target object in the input image has one corresponding candidate frame set; determine a predetermined number of candidate frames with overlap degrees thereof ranked in the top as reference frames; generating a size-related feature according to respective sizes of the reference frames and a size of the enclosing frame; and detect an object in the input image by applying the size-related feature in a machine learning model. ### Claim 7 (depends on 6) The object detection device according to claim 6, wherein the one or more programs are executed by the one or more processors to enable the one or more processors to: for each candidate frame, align the enclosing frame with the candidate frame at one vertex; calculate an intersection-over-union of the enclosing frame with the candidate frame; and determine the calculated intersection-over-union as the overlap degree of the enclosing frame with the candidate frame. ### Claim 8 (depends on 6) The object detection device according to claim 6, wherein the one or more programs are executed by the one or more processors to enable the one or more processors to: in a case that the overlap degree of the enclosing frame with the candidate frame is greater than an overlap threshold, determine the candidate frame as the reference frame. ### Claim 9 (depends on 6) The object detection device according to claim 6, wherein the one or more programs are executed by the one or more processors to enable the one or more processors to: determine a width ratio and a height ratio of the enclosing frame with the reference frame as the size-related feature, or determine a natural logarithm of the width ratio and a natural logarithm of the height ratio as the size-related feature. ### Claim 10 (depends on 6) The object detection device according to claim 6, wherein the one or more programs are executed by the one or more processors to enable the one or more processors to: predetermine a relationship between respective candidate frame sets and respective object categories; determine an object category to which the object to be detected belongs; and select a candidate frame set according to the determined object category. ### Claim 11 (depends on 1) A non-transitory computer-readable storage medium, in which a computer program is stored, wherein the computer program, when executed by a processor, causes the processor to implement the method of claim 1.