CN111368878B - Optimization method based on SSD target detection, computer equipment and medium - Google Patents
Optimization method based on SSD target detection, computer equipment and medium Download PDFInfo
- Publication number
- CN111368878B CN111368878B CN202010093422.3A CN202010093422A CN111368878B CN 111368878 B CN111368878 B CN 111368878B CN 202010093422 A CN202010093422 A CN 202010093422A CN 111368878 B CN111368878 B CN 111368878B
- Authority
- CN
- China
- Prior art keywords
- negative sample
- sample candidate
- frame
- loss function
- box
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Abstract
The invention discloses an optimization method based on SSD target detection, a computer readable storage medium and computer equipment, wherein the optimization method comprises the following steps: inputting a training image into a target detection network and acquiring a plurality of candidate frames; respectively comparing each candidate frame with a preset sample truth frame to obtain a positive sample candidate frame and a negative sample candidate frame; respectively calculating the classification probability of each negative sample candidate frame; calculating a first loss function value and a second loss function value of each negative sample candidate frame according to the classification probability of each negative sample candidate frame; acquiring a preset number of first negative sample candidate frames according to the first loss function values of the negative sample candidate frames; acquiring a preset number of second negative sample candidate frames according to the second loss function values of the negative sample candidate frames; optimizing the target detection network according to the positive sample candidate box, the first negative sample candidate box, and the second negative sample candidate box. The embodiment provided by the invention can effectively reduce the false detection rate of the classification loss function of the target detection network.
Description
Technical Field
The present invention relates to the field of SSD target detection, and in particular, to an optimization method based on SSD target detection, a computer-readable storage medium, and a computer device.
Background
With the rapid development of the third wave artificial intelligence, various optimization methods based on the deep learning target detection network continuously appear. In the target detection application, the loss function of target detection generally includes a positioning loss function and a classification loss function, and the classification loss function in the prior art mostly adopts a softmax loss function and a focalls loss function. However, in practical application, it is found that a certain proportion of false detection rate exists in the existing target detection network, and the accuracy rate of target detection is reduced.
Disclosure of Invention
In order to solve the above problem, a first embodiment of the present invention provides an optimization method based on SSD target detection, including:
inputting a training image into an SSD target detection network and obtaining a plurality of candidate frames;
respectively comparing each candidate frame with a preset sample truth frame to obtain a positive sample candidate frame and a negative sample candidate frame;
respectively calculating the classification probability of each negative sample candidate frame;
calculating a first loss function value and a second loss function value of each negative sample candidate frame according to the classification probability of the negative sample candidate frame;
acquiring a preset number of first negative sample candidate frames according to the first loss function values of the negative sample candidate frames;
acquiring a preset number of second negative sample candidate frames according to the second loss function values of the negative sample candidate frames;
optimizing the SSD target detection network using a gradient optimization method according to the positive sample candidate box, the first negative sample candidate box, and the second negative sample candidate box.
Further, the classification probability of the negative sample candidate box is:
wherein j is a classification category, a j Sample input value for j classification category, T is classification number, a k The values are input for the samples of each classification category.
Further, the first loss function value of the negative sample candidate box is:
wherein j is a classification category, y j Representing whether the negative sample candidate box belongs to the category j, if so, y j =1, otherwise y j ＝0。
Further, the second loss function value of the negative sample candidate box is:
L′＝max(S j )-S 0
wherein S is j Is the classification probability of the negative sample candidate box j>0，S 0 And the classification probability of the negative sample candidate box as a preset category is obtained.
Further, the obtaining a preset number of first negative sample candidate frames according to the first loss function value of the negative sample candidate frame further includes:
NMS screening is carried out on each negative sample candidate box so as to delete redundant negative sample candidate boxes;
and sorting according to the first loss function values of the screened negative sample candidate frames, and selecting a preset number of first negative sample candidate frames according to a sorting sequence.
Further, the obtaining a preset number of second negative sample candidate frames according to the second loss function value of the negative sample candidate frame further includes:
NMS screening is carried out on each negative sample candidate box so as to delete redundant negative sample candidate boxes;
and sorting according to the second loss function values of the screened negative sample candidate frames and selecting a preset number of second negative sample candidate frames according to the sorting sequence.
Further, the respectively comparing each candidate box with a preset sample truth box to obtain a positive sample candidate box and a negative sample candidate box further includes:
comparing each candidate frame with a preset sample truth frame respectively and outputting a comparison result;
and judging whether the comparison result is greater than a preset coincidence threshold, if so, determining the frame as a positive sample candidate frame, otherwise, determining the frame as a negative sample candidate frame.
Further, the ratio of the number of the positive sample candidate boxes, the number of the first negative sample candidate boxes, and the number of the second negative sample candidate boxes is: 1:1.5:1.5.
A second embodiment of the invention provides a computer-readable storage medium, on which a computer program is stored, which program, when being executed by a processor, carries out the method according to the first embodiment.
A third embodiment of the invention provides a computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method according to the first embodiment when executing the program.
The invention has the following beneficial effects:
aiming at the existing problems, the invention sets an optimization method based on SSD target detection, a computer readable storage medium and a computer device, obtains a first negative sample candidate frame and a first negative sample candidate frame by calculating the classification probability, a first loss function value and a second loss function value of each negative sample candidate frame, and optimizes the SSD target detection network by using a gradient optimization method according to the positive sample candidate frame, the first negative sample candidate frame and the second negative sample candidate frame, thereby solving the problems in the prior art, effectively reducing the false detection rate of the classification loss function, improving the recall rate and the accuracy rate of the SSD target detection network, and having wide application prospect.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.
FIG. 1 illustrates a flow diagram of an optimization method according to an embodiment of the invention;
FIG. 2 shows a schematic diagram of a training image according to an embodiment of the present invention;
fig. 3 is a schematic structural diagram of a computer device according to another embodiment of the present invention.
Detailed Description
In order to more clearly illustrate the invention, the invention is further described below with reference to preferred embodiments and the accompanying drawings. Similar components in the figures are denoted by the same reference numerals. It is to be understood by persons skilled in the art that the following detailed description is illustrative and not restrictive, and is not to be taken as limiting the scope of the invention.
As shown in fig. 1, an embodiment of the present invention provides an optimization method based on SSD target detection, including: inputting a training image into an SSD target detection network and obtaining a plurality of candidate frames; respectively comparing each candidate frame with a preset sample truth frame to obtain a positive sample candidate frame and a negative sample candidate frame; respectively calculating the classification probability of each negative sample candidate frame; calculating a first loss function value and a second loss function value of each negative sample candidate frame according to the classification probability of the negative sample candidate frame; acquiring a preset number of first negative sample candidate frames according to the first loss function values of the negative sample candidate frames; acquiring a preset number of second negative sample candidate frames according to the second loss function values of the negative sample candidate frames; optimizing the SSD target detection network using a gradient optimization method according to the positive sample candidate box, the first negative sample candidate box, and the second negative sample candidate box.
In a specific example, as shown in fig. 2, the target detection is performed on the training image by the following steps:
the first step is to input a training image into the SSD target detection network and obtain a plurality of candidate boxes.
In the present embodiment, as shown in fig. 2, a training image generates a large number of candidate frames via an SSD object detection network, and performs object recognition and classification on each candidate frame.
It should be noted that the present application may also process the video, for example, cut the video into a plurality of image frames, and input the image frames into the SSD object detection network for object detection.
And secondly, comparing each candidate frame with a preset sample truth frame to obtain a positive sample candidate frame and a negative sample candidate frame.
In this embodiment, a sample truth image including a detection target is input to the SSD target detection network in advance, a sample truth frame is obtained according to the sample truth image, and the obtained candidate frames are compared with a sample truth frame of a preset sample truth image, so as to obtain a positive sample candidate frame and a negative sample candidate frame.
In an optional embodiment, the method specifically includes:
and respectively comparing each candidate box with a preset sample truth box and outputting a comparison result.
In this embodiment, each candidate frame is compared with the sample truth frame, and the similarity between the candidate frame and the sample truth frame is calculated.
And judging whether the comparison result is greater than a preset coincidence threshold, if so, determining the frame as a positive sample candidate frame, otherwise, determining the frame as a negative sample candidate frame.
In this embodiment, the coincidence threshold is 0.4, and if the coincidence degree between the candidate frame and the sample true value frame is greater than 0.4, the candidate frame is set as a positive sample candidate frame, for example, if the similarity between the candidate frame 1 shown in fig. 2 and the sample true value frame of the preset "dog" is greater than 0.4, the candidate frame is determined as a positive sample candidate frame; otherwise, if the similarity between the candidate frame and the sample true value frame is less than 0.4, the candidate frame is set as a negative sample candidate frame, for example, if the similarity between the candidate frame 2 shown in fig. 2 and the sample true value frame of the preset "dog" is less than 0.4, the candidate frame is determined as a negative sample candidate frame, for example, if the overlap between the candidate frame 3 shown in fig. 2 and the sample true value frame of the preset "dog" is less than 0.4, the candidate frame is determined as a negative sample candidate frame.
It should be noted that, the larger the coincidence threshold is, the higher the accuracy of target detection is, but the number of candidate frames of the positive sample is also reduced, and those skilled in the art should select an appropriate coincidence threshold according to the actual application requirements to implement target detection as a design criterion, which is not described herein again.
And thirdly, respectively calculating the classification probability of each negative sample candidate frame.
In this embodiment, the obtained negative sample candidate box is processed, and in consideration of using the negative sample candidate box for a training task, if the negative sample candidate box is too many, a loss function of the target detection network is too large, which causes a false detection rate, and reduces a recall rate and an accuracy rate of the target detection network.
To avoid this problem, the classification probability of each negative sample candidate box is calculated, specifically:
wherein j is a classification category, a j Sample input values of j classification categories, T is the number of classifications, j ranges from 0 to T, a k The value is input for the sample for each classification category.
In this embodiment, the classification categories include "background", "dog", "cat", "chicken" and "rabbit", i.e. T is 5, j is "background", "dog", "cat", "chicken" and "rabbit", respectively, and the classification probabilities S of candidate boxes 2 shown in fig. 2 are respectively j Are [0.2,0.3, 0.1 ]]Then the classification probabilities S of the candidate boxes 3 shown in FIG. 2 j Are [0.29,0.59,0.1,0.01 ]]。
And fourthly, calculating a first loss function value and a second loss function value of each negative sample candidate frame according to the classification probability of each negative sample candidate frame.
In this embodiment, for the calculated classification probability of each negative sample candidate frame, the first loss function value of each negative sample candidate frame is calculated using the first loss function, respectively.
Specifically, calculating a first loss function value softmax of each negative sample candidate box by a first loss function softmax:
wherein j is a classification category, y j Representing whether the negative sample candidate box belongs to the category j, if so, y j =1, otherwise y j ＝0。
In this embodiment, the first loss function value of the frame candidate 2 shown in fig. 2 for "background" is-log 0.2, and the first loss function value of the frame candidate 3 shown in fig. 2 for "background" is-log 0.29, and the first loss function value of the frame candidate 2 is larger than the first loss function value of the frame candidate 3.
In this embodiment, for the calculated classification probability of each negative sample candidate frame, the second loss function value of each negative sample candidate frame is calculated using the second loss function, respectively.
Specifically, calculating a second loss function value softmax inter-class loss value of each negative sample candidate frame by using a second loss function softmax:
L′＝max(S j )-S 0
wherein S is j Is the classification probability, S, of the negative sample candidate box 0 And classifying the probability that the negative sample candidate box is in a preset category.
In this embodiment, the preset category is a background category of the negative sample category, i.e. S 0 The classification probability of the background class of the negative sample candidate box. Specifically, the second loss function value of the candidate frame 2 shown in fig. 2 as the "background" is 0.1 (i.e., 0.3-0.2), and the second loss function value of the candidate frame 3 shown in fig. 2 as the "background" is 0.2 (i.e., 0.5-0.3, then the second loss function value of the candidate frame 3 is greater than the first loss function value of the candidate frame 2, so that by calculating the negative sample candidate frame using the second loss function, the candidate frame 3 is selected to optimize the neural network parameter with a higher inter-class loss (second class loss function value), which can effectively reduce the false detection rate of the classification loss function, thereby improving the recall rate and accuracy of the target detection.
And fifthly, acquiring a preset number of first negative sample candidate frames according to the first loss function values of the negative sample candidate frames.
In this embodiment, obtaining the first negative sample candidate box through the first loss function value specifically includes:
and performing NMS screening on each negative sample candidate box to delete redundant negative sample candidate boxes.
In this embodiment, each negative sample candidate box is screened through an NMS operation (non-maximum suppression), that is, a negative sample box with an excessively high degree of coincidence in the negative sample candidate box is deleted, and only one negative sample candidate box is retained, so as to optimize the SSD target detection network.
And sorting according to the first loss function values of the screened negative sample candidate frames, and selecting a preset number of first negative sample candidate frames according to a sorting sequence.
In this embodiment, each of the negative sample candidate frames remaining after the screening is sorted according to the first loss function value, and a preset number of negative sample candidate frames sorted in the top are selected as the first negative sample candidate frame.
And sixthly, acquiring a preset number of second negative sample candidate frames according to the second loss function values of the negative sample candidate frames.
In this embodiment, the obtaining the first negative sample candidate box through the second loss function value at the same time specifically includes:
and performing NMS screening on each negative sample candidate box to delete redundant negative sample candidate boxes.
In this embodiment, each negative sample candidate box is screened by NMS operation (non-maximum suppression), that is, a negative sample box with a high degree of coincidence in the negative sample candidate boxes is deleted, and only one negative sample candidate box is reserved, so as to optimize the SSD target detection network.
And sorting according to the second loss function values of the screened negative sample candidate frames and selecting a preset number of second negative sample candidate frames according to the sorting sequence.
In this embodiment, each of the negative sample candidate frames remaining after the screening is sorted according to the second loss function value, and a preset number of negative sample candidate frames sorted in the top are selected as the second negative sample candidate frames.
It should be noted that, in consideration of the number of positive sample candidate boxes selected, a negative sample candidate box is selected according to a ratio of 1. Those skilled in the art should select an appropriate ratio according to the actual application requirement to achieve the purpose of target detection as a design criterion, and details are not described herein.
Finally, the SSD target detection network is optimized using a gradient optimization method according to the positive sample candidate box, the first negative sample candidate box, and the second negative sample candidate box.
In this example, the mini-SGD is selected for gradient optimization.
So far, SSD target detection of the training image is completed.
Another embodiment of the present invention provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements: inputting a training image into an SSD target detection network and obtaining a plurality of candidate frames; respectively comparing each candidate frame with a preset sample truth frame to obtain a positive sample candidate frame and a negative sample candidate frame; respectively calculating the classification probability of each negative sample candidate frame; calculating a first loss function value and a second loss function value of each negative sample candidate frame according to the classification probability of the negative sample candidate frame; acquiring a preset number of first negative sample candidate frames according to the first loss function values of the negative sample candidate frames; acquiring a preset number of second negative sample candidate frames according to the second loss function values of the negative sample candidate frames; optimizing the SSD target detection network using a gradient optimization method according to the positive sample candidate box, the first negative sample candidate box, and the second negative sample candidate box.
In practice, the computer-readable storage medium may take any combination of one or more computer-readable media. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present embodiment, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device.
A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take many forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to wireless, wireline, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
Computer program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the case of a remote computer, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider).
As shown in fig. 3, another embodiment of the present invention provides a schematic structural diagram of a computer device. The computer device 12 shown in FIG. 3 is only an example and should not impose any limitation on the scope of use or functionality of embodiments of the present invention.
As shown in FIG. 3, computer device 12 is in the form of a general purpose computing device. The components of computer device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 that couples various system components including the system memory 28 and the processing unit 16.
Computer device 12 typically includes a variety of computer system readable media. Such media can be any available media that is accessible by computer device 12 and includes both volatile and nonvolatile media, removable and non-removable media.
The system memory 28 may include computer system readable media in the form of volatile memory, such as Random Access Memory (RAM) 30 and/or cache memory 32. Computer device 12 may further include other removable/non-removable, volatile/nonvolatile computer system storage media. By way of example only, storage system 34 may be used to read from and write to non-removable, nonvolatile magnetic media (not shown in FIG. 3, and commonly referred to as a "hard drive"). Although not shown in FIG. 3, a magnetic disk drive for reading from and writing to a removable, nonvolatile magnetic disk (e.g., a "floppy disk") and an optical disk drive for reading from or writing to a removable, nonvolatile optical disk (e.g., a CD-ROM, DVD-ROM, or other optical media) may be provided. In these cases, each drive may be connected to bus 18 by one or more data media interfaces. Memory 28 may include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the invention.
A program/utility 40 having a set (at least one) of program modules 42 may be stored, for example, in memory 28, such program modules 42 including, but not limited to, an operating system, one or more application programs, other program modules, and program data, each of which examples or some combination thereof may comprise an implementation of a network environment. Program modules 42 generally carry out the functions and/or methodologies of embodiments of the invention as described.
Computer device 12 may also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), with one or more devices that enable a user to interact with computer device 12, and/or with any devices (e.g., network card, modem, etc.) that enable computer device 12 to communicate with one or more other computing devices. Such communication may be through an input/output (I/O) interface 22. Also, computer device 12 may communicate with one or more networks (e.g., a Local Area Network (LAN), a Wide Area Network (WAN), and/or a public network such as the Internet) via network adapter 20. As shown in FIG. 3, the network adapter 20 communicates with the other modules of the computer device 12 via the bus 18. It should be understood that although not shown in FIG. 3, other hardware and/or software modules may be used in conjunction with computer device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, to name a few.
The processor unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, for example, to implement an optimization method based on SSD object detection provided by the embodiments of the present invention.
Aiming at the existing problems, the invention sets an optimization method based on SSD target detection, a computer readable storage medium and a computer device, obtains a first negative sample candidate frame and a first negative sample candidate frame by calculating the classification probability, a first loss function value and a second loss function value of each negative sample candidate frame, and optimizes the SSD target detection network by using a gradient optimization method according to the first loss function value and the second loss function value of each negative sample candidate frame, thereby making up the problems in the prior art, effectively reducing the false detection rate of the classification loss function, improving the recall rate and the accuracy rate of the SSD target detection network, and having wide application prospects.
It should be understood that the above-mentioned embodiments of the present invention are only examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention, and it will be obvious to those skilled in the art that other variations or modifications may be made on the basis of the above description, and all embodiments may not be exhaustive, and all obvious variations or modifications may be included within the scope of the present invention.
Claims (7)
1. An optimization method based on SSD target detection is characterized by comprising the following steps:
inputting a training image into an SSD target detection network and obtaining a plurality of candidate frames;
respectively comparing each candidate frame with a preset sample truth frame to obtain a positive sample candidate frame and a negative sample candidate frame;
respectively calculating the classification probability of each negative sample candidate frame;
calculating a first loss function value and a second loss function value of each negative sample candidate frame according to the classification probability of the negative sample candidate frame;
acquiring a preset number of first negative sample candidate frames according to the first loss function values of the negative sample candidate frames;
acquiring a preset number of second negative sample candidate frames according to the second loss function values of the negative sample candidate frames;
optimizing the SSD target detection network using a gradient optimization method according to the positive sample candidate box, the first negative sample candidate box, and the second negative sample candidate box;
the classification probability of the negative sample candidate box is as follows:
wherein j is a classification category, a j For the sample input value of j classification category, T is the number of classifications, a k Inputting values for the samples of each classification category;
the first loss function value for the negative sample candidate box is:
wherein j is a classification category, y j Representing whether the negative sample candidate box belongs to the category j, if so, y j =1, otherwise y j ＝0；
The second loss function value for the negative sample candidate box is:
L′＝max(S j )-S 0
wherein S is j Is the classification probability, j, of the negative sample candidate box>0，S 0 And classifying the probability that the negative sample candidate box is in a preset category.
2. The optimization method according to claim 1, wherein the obtaining a preset number of first negative sample candidate boxes according to the first loss function values of the negative sample candidate boxes further comprises:
NMS screening is carried out on each negative sample candidate box so as to delete redundant negative sample candidate boxes;
and sorting according to the first loss function values of the screened negative sample candidate frames and selecting a preset number of first negative sample candidate frames according to the sorting sequence.
3. The optimization method according to claim 1, wherein the obtaining a preset number of second negative sample candidate boxes according to the second loss function values of the negative sample candidate boxes further comprises:
NMS screening is carried out on each negative sample candidate box so as to delete redundant negative sample candidate boxes;
and sorting according to the second loss function values of the screened negative sample candidate frames and selecting a preset number of second negative sample candidate frames according to the sorting sequence.
4. The optimization method of claim 1, wherein the comparing each candidate box with a preset sample truth box to obtain a positive sample candidate box and a negative sample candidate box further comprises:
respectively comparing each candidate frame with a preset sample truth value frame and outputting a comparison result;
and judging whether the comparison result is greater than a preset coincidence threshold, if so, determining the frame as a positive sample candidate frame, otherwise, determining the frame as a negative sample candidate frame.
5. The optimization method according to claim 1, wherein the ratio of the number of positive sample candidate boxes, the number of first negative sample candidate boxes and the number of second negative sample candidate boxes is: 1:1.5:1.5.
6. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the method according to any one of claims 1-5.
7. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the method according to any of claims 1-5 when executing the program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010093422.3A CN111368878B (en) | 2020-02-14 | 2020-02-14 | Optimization method based on SSD target detection, computer equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010093422.3A CN111368878B (en) | 2020-02-14 | 2020-02-14 | Optimization method based on SSD target detection, computer equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111368878A CN111368878A (en) | 2020-07-03 |
| CN111368878B true CN111368878B (en) | 2023-02-28 |
Family
ID=71212495
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010093422.3A Active CN111368878B (en) | 2020-02-14 | 2020-02-14 | Optimization method based on SSD target detection, computer equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111368878B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112507787B (en) * | 2020-11-03 | 2022-08-05 | 北京深睿博联科技有限责任公司 | Target detection method and device for screening visual barrier personnel obstacles based on positive sample |
| CN112766181B (en) * | 2021-01-22 | 2022-09-23 | 电子科技大学 | Method for improving line graph identification accuracy |
| CN112906823B (en) * | 2021-03-29 | 2022-07-05 | 苏州科达科技股份有限公司 | Target object recognition model training method, recognition method and recognition device |
| CN116310656B (en) * | 2023-05-11 | 2023-08-15 | 福瑞泰克智能系统有限公司 | Training sample determining method and device and computer equipment |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108304820A (en) * | 2018-02-12 | 2018-07-20 | 腾讯科技（深圳）有限公司 | A kind of method for detecting human face, device and terminal device |
| WO2018137357A1 (en) * | 2017-01-24 | 2018-08-02 | 北京大学 | Target detection performance optimization method |
| CN109409252A (en) * | 2018-10-09 | 2019-03-01 | 杭州电子科技大学 | A kind of traffic multi-target detection method based on modified SSD network |
| CN109919251A (en) * | 2019-03-21 | 2019-06-21 | 腾讯科技（深圳）有限公司 | A kind of method and device of object detection method based on image, model training |
| CN110533022A (en) * | 2019-06-11 | 2019-12-03 | 广州海昇教育科技有限责任公司 | A kind of object detection method, system, device and storage medium |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7756313B2 (en) * | 2005-11-14 | 2010-07-13 | Siemens Medical Solutions Usa, Inc. | System and method for computer aided detection via asymmetric cascade of sparse linear classifiers |
-
2020
- 2020-02-14 CN CN202010093422.3A patent/CN111368878B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018137357A1 (en) * | 2017-01-24 | 2018-08-02 | 北京大学 | Target detection performance optimization method |
| CN108304820A (en) * | 2018-02-12 | 2018-07-20 | 腾讯科技（深圳）有限公司 | A kind of method for detecting human face, device and terminal device |
| CN109409252A (en) * | 2018-10-09 | 2019-03-01 | 杭州电子科技大学 | A kind of traffic multi-target detection method based on modified SSD network |
| CN109919251A (en) * | 2019-03-21 | 2019-06-21 | 腾讯科技（深圳）有限公司 | A kind of method and device of object detection method based on image, model training |
| CN110533022A (en) * | 2019-06-11 | 2019-12-03 | 广州海昇教育科技有限责任公司 | A kind of object detection method, system, device and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 基于卷积神经网络的实时行人检测方法;龚安等;《计算机系统应用》;20170915(第09期);第219-222页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111368878A (en) | 2020-07-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111368878B (en) | Optimization method based on SSD target detection, computer equipment and medium | |
| CN109614625B (en) | Method, device and equipment for determining title text relevancy and storage medium | |
| US11640551B2 (en) | Method and apparatus for recommending sample data | |
| CN110942011B (en) | Video event identification method, system, electronic equipment and medium | |
| CN112906823B (en) | Target object recognition model training method, recognition method and recognition device | |
| CN113434683B (en) | Text classification method, device, medium and electronic equipment | |
| CN110188766B (en) | Image main target detection method and device based on convolutional neural network | |
| EP3620982A1 (en) | Sample processing method and device | |
| CN111753863A (en) | Image classification method and device, electronic equipment and storage medium | |
| CN112149754B (en) | Information classification method, device, equipment and storage medium | |
| CN110781849A (en) | Image processing method, device, equipment and storage medium | |
| CN113486178B (en) | Text recognition model training method, text recognition method, device and medium | |
| CN113239883A (en) | Method and device for training classification model, electronic equipment and storage medium | |
| CN113936232A (en) | Screen fragmentation identification method, device, equipment and storage medium | |
| CN113782029A (en) | Training method, device and equipment of speech recognition model and storage medium | |
| CN111738290B (en) | Image detection method, model construction and training method, device, equipment and medium | |
| CN111259932A (en) | Classification method, medium, device and computing equipment | |
| CN113033817B (en) | OOD detection method and device based on hidden space, server and storage medium | |
| CN110442714B (en) | POI name normative evaluation method, device, equipment and storage medium | |
| CN110059180B (en) | Article author identity recognition and evaluation model training method and device and storage medium | |
| CN113762308A (en) | Training method, classification method, device, medium and equipment of classification model | |
| CN114065915A (en) | Network model construction method, data processing method, device, medium and equipment | |
| CN113032564B (en) | Feature extraction method, device, electronic equipment and storage medium | |
| CN116166920A (en) | Business object analysis method, device, equipment and storage medium | |
| CN114782954A (en) | Picture identifying code identification method, system, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |