CN108256490A - Method for detecting human face and listen to the teacher rate appraisal procedure, system based on Face datection - Google Patents
Method for detecting human face and listen to the teacher rate appraisal procedure, system based on Face datection Download PDFInfo
- Publication number
- CN108256490A CN108256490A CN201810073494.4A CN201810073494A CN108256490A CN 108256490 A CN108256490 A CN 108256490A CN 201810073494 A CN201810073494 A CN 201810073494A CN 108256490 A CN108256490 A CN 108256490A
- Authority
- CN
- China
- Prior art keywords
- face
- teacher
- rate
- detected
- listening
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 83
- 230000001815 facial effect Effects 0.000 claims abstract description 20
- 206010061274 Malocclusion Diseases 0.000 claims abstract description 13
- 238000012937 correction Methods 0.000 claims abstract description 8
- 238000001514 detection method Methods 0.000 claims description 30
- 238000010586 diagram Methods 0.000 claims description 12
- 238000012216 screening Methods 0.000 claims description 9
- 230000004044 response Effects 0.000 claims description 4
- 238000003709 image segmentation Methods 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 description 12
- 238000005516 engineering process Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 235000007926 Craterellus fallax Nutrition 0.000 description 2
- 240000007175 Datura inoxia Species 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000000295 complement effect Effects 0.000 description 2
- 238000007418 data mining Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 239000000203 mixture Substances 0.000 description 2
- 230000008901 benefit Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000000977 initiatory effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/161—Detection; Localisation; Normalisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/20—Education
- G06Q50/205—Education administration or guidance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Educational Administration (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Oral & Maxillofacial Surgery (AREA)
- Educational Technology (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Marketing (AREA)
- Primary Health Care (AREA)
- General Business, Economics & Management (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The present invention discloses a kind of method for detecting human face and listen to the teacher rate appraisal procedure, system based on Face datection, the method for detecting human face include:Using classroom picture to be detected as input, object candidate window is generated using selective search method for step 1;Object candidate window is input in the combination face characteristic grader of multiple parallel connections and is further screened by step 2, obtains face candidate window;Face candidate window is input in a multitask depth convolutional network by step 3, carries out further Face datection and window correction, and the face to detecting carries out angle classification;Step 4 is listened to the teacher the detected value of rate according to what facial angle was classified as a result, obtaining classroom in picture to be detected, the present invention can be achieved a kind of high-performance for being suitble to apply in complex environment, high-accuracy, easily realize, rate assessment system of listening to the teacher.
Description
Technical field
The present invention relates to deep learning and computer vision field, more particularly to a kind of method for detecting human face and based on people
Listen to the teacher rate appraisal procedure, the system of face detection.
Background technology
As colleges and universities pay attention to and data mining technology flourished in recent years the continuous of teachers ' teaching quality, at present not
Few colleges and universities propose that the modern times comment teaching system one after another.It utilizes data mining technology, and from the data commented in teaching system, (such as teacher's course is pacified
Row's resonable degree, operation layout resonable degree, course content difficulty or ease resonable degree, rate of listening to the teacher etc.) effective information is obtained, with
Assessment further is made to the quality of instruction of teacher.And rate assessment of listening to the teacher at present relies primarily on manpower to obtain, and on the one hand causes
The waste of human resources, on the other hand causes to introduce excessive artificial subjective factor, so as to comment the bottle of teaching system as the modern times
Neck.
The listen to the teacher essence of rate assessment of classroom is that student listens to the teacher the statistics of number, can be obtained by human face detection tech
It takes.There is the detection Shandongs under Detection accuracy difference and serious circumstance of occlusion under complex environment for the human face detection tech of mainstream at present
Stick is poor.And the human face detection tech based on deep learning can obtain the feature of more horn of plenty, and since its training is without prison
Superintending and directing property so that it is more suitable in complex environment and applies.
Therefore, how to design a kind of high-performance for being suitble to apply in complex environment, high-accuracy, easily realize, listen to the teacher
Rate assessment system is conducive to the quantizating index for teaching system being commented to provide the rate of listening to the teacher for the modern times, is that school comments teaching system suddenly to wait to solve at present
One of certainly the problem of, the present invention for it is existing listen to the teacher rate assessment strategy the problem of, it is proposed that a kind of method for detecting human face
And listen to the teacher rate appraisal procedure, system based on Face datection.
Invention content
To overcome above-mentioned the shortcomings of the prior art, the present invention's is designed to provide a kind of method for detecting human face and base
Rate appraisal procedure of listening to the teacher, system in Face datection, using picture to be detected or video as input, utilize face according to demand
Detection method carries out input Face datection and facial angle is classified, and obtains the detected value for rate of listening to the teacher, then utilizes demographics
Algorithm carries out input the estimation of total number of persons, the detected value for rate of finally being listened to the teacher according to classroom and the estimated value of total number of persons, using listening
Class rate calibration method obtains final rate assessed value of listening to the teacher.
In view of the above and other objects, the present invention proposes a kind of method for detecting human face, include the following steps:
Using classroom picture to be detected as input, object candidate window is generated using selective search method for step 1;
Object candidate window is input in the combination face characteristic grader of multiple parallel connections and carries out further by step 2
Screening obtains face candidate window;
Face candidate window is input in a multitask depth convolutional network, carries out further face by step 3
Detection and window correction, and the face to detecting carries out angle classification
Step 4 is listened to the teacher the detected value of rate according to what facial angle was classified as a result, obtaining classroom in picture to be detected.
Further, step 1 further comprises:
The target object of acquisition arbitrary dimension is removed using region subdivision grouping algorithm;
All situations that may be present are handled using diversified feature policy;
Combining position information, the region obtained by region subdivision grouping algorithm and diversified Image Segmentation Methods Based on Features strategy, selection
Property search screening is ranked up come the region to generation using location information, filter out the region sorted rearward.
Further, in step 2, the object candidate window obtained from step 1 is input to four based on depth convolution
It is screened in the combination face characteristic grader of network, after obtaining face candidate window, utilizes non-maximum value restrainable algorithms pair
Face candidate window carries out window duplicate removal processing.
Further, four assemblage characteristic graders share identical depth convolutional network structure, the network knot
Structure includes 7 convolutional layers and 3 pond layers.
Further, in step 4, the facial angle classification results and Face datection obtained according to step 3 will be as a result, will
As classroom total number of persons, the former described the latter listen to the teacher total number of persons as classroom, and the ratio of the two is then listened to the teacher for specific time period classroom
The detected value of rate.
Further, it before step 4, further includes:
By picture to be detected be input to a multitask depth convolutional network as a result, obtain picture in total number of persons estimation
Value, and using the value as the estimated value of a certain moment classroom total number of persons.
Further, the multitask depth convolutional network includes 3 convolutional layers and 2 pond layers, and multitask includes people
Population density response diagram and Population size estimation.
In order to achieve the above objectives, the present invention also provides a kind of face detection system, including:
Object candidate window generation unit, for using classroom picture to be detected as input, utilizing selective search side
Method generates object candidate window;
Face candidate window acquiring unit, for object candidate window to be input to the combination face characteristic point of multiple parallel connections
It is further screened in class device, obtains face candidate window;
Facial angle taxon, for face candidate window to be input in a multitask depth convolutional network, into
The further Face datection of row and window correction, and the face to detecting carries out angle classification
Rate of listening to the teacher detected value computing unit, for according to facial angle classify as a result, obtaining classroom in picture to be detected
The detected value for rate of listening to the teacher.
In order to achieve the above objectives, the present invention also provides a kind of rate appraisal procedure of listening to the teacher based on Face datection, including as follows
Step:
Step 1, according to the rate detected value of listening to the teacher that method for detecting human face obtains, according to a predetermined threshold value to the detected value
It is screened, deletes improper data;
Step 2, the detection of listen to the teacher total number of persons estimated value and method for detecting human face acquisition obtained according to method for detecting human face
Value obtains the number not detected in classroom;
Step 3 according to the rate detected value of listening to the teacher that above-mentioned method for detecting human face obtains, obtains its minimum as step 2
The rate of listening to the teacher for not detecting number, the calibration for rate of being listened to the teacher with completion.
In order to achieve the above objectives, the present invention also provides a kind of rate assessment system of listening to the teacher based on Face datection, including:
Detected value screening unit, for the rate detected value of listening to the teacher obtained according to method for detecting human face, according to a predetermined threshold value
The detected value is screened, deletes improper data;
Number acquiring unit is not detected, for listen to the teacher total number of persons estimated value and the face inspection obtained according to method for detecting human face
The detected value that survey method obtains obtains the number not detected in classroom;
Alignment unit for the rate detected value of listening to the teacher obtained according to above-mentioned method for detecting human face, obtains its minimum conduct
The rate of listening to the teacher for not detecting number described in number acquiring unit is not detected, the calibration for rate of listening to the teacher with completion.
Compared with prior art, a kind of method for detecting human face of the present invention and rate appraisal procedure of listening to the teacher based on Face datection,
System according to demand using picture to be detected or video as input, input is carried out using method for detecting human face Face datection with
Facial angle is classified, and obtains the detected value for rate of listening to the teacher, and then carries out the estimation of total number of persons to input using people counting algorithm, most
The detected value for rate of being listened to the teacher eventually according to classroom and the estimated value of total number of persons, final rate of listening to the teacher is obtained using rate calibration method is listened to the teacher
Assessed value, inventive energy is good, accuracy rate is high, easy realization, is conducive to the quantizating index for teaching system being commented to provide the rate of listening to the teacher for the modern times.
Description of the drawings
Fig. 1 is a kind of step flow chart of method for detecting human face of the present invention;
Fig. 2 is the flow diagram of the method for detecting human face of the specific embodiment of the invention;
Fig. 3 is the depth convolutional network structure chart of an assemblage characteristic grader in the specific embodiment of the invention;
Fig. 4 is the structure chart of a multitask depth convolutional network in the specific embodiment of the invention;
Fig. 5 is the schematic diagram of another multitask depth convolutional network structure in the specific embodiment of the invention;
Fig. 6 is a kind of system architecture diagram of face detection system of the present invention;
Fig. 7 is a kind of step flow chart of the rate assessment of listening to the teacher based on Face datection of the present invention;
Fig. 8 is a kind of system architecture diagram of the rate assessment system of listening to the teacher based on Face datection of the present invention.
Specific embodiment
Below by way of specific specific example and embodiments of the present invention are described with reference to the drawings, those skilled in the art can
Understand the further advantage and effect of the present invention easily by content disclosed in the present specification.The present invention can also pass through other differences
Specific example implemented or applied, the various details in this specification also can be based on different viewpoints with application, without departing substantially from
Various modifications and change are carried out under the spirit of the present invention.
Fig. 1 is a kind of step flow chart of method for detecting human face of the present invention, and Fig. 2 is that the face of the specific embodiment of the invention is examined
The flow diagram of survey method.As shown in Figures 1 and 2, a kind of method for detecting human face of the present invention, includes the following steps:
Step 101, using classroom picture to be detected as input, object candidate's window is generated using selective search method
Mouthful.
Specifically, in embodiments of the present invention, by the use of selective search as generation object candidate window method, first
The target object of acquisition arbitrary dimension is removed using region subdivision grouping algorithm, algorithm thinks that the information that region includes will be far more than list
The information that pixel is included, thus the feature based on region is used as far as possible.When obtaining a series of small initiation regions, adopt
With fast area partitioning algorithm, after having obtained required prime area, the algorithm is using greedy algorithm by continuous iteration to area
Domain is combined, to obtain required regional ensemble;Next employs diversified feature policy and there may be to handle
Situation, mainly include three kinds of feature policies, be color space diversification respectively, complementary similarity calculating method, complementary
Initialization area, to cover region as much as possible;It is finally combining position information, by region subdivision grouping algorithm and various
The region that change Image Segmentation Methods Based on Features strategy obtains is excessive, and selective search is ranked up sieve using location information come the region to generation
Choosing filters out the region of sequence rearward, and the region ordering for ensureing to repeat to be labeled can be forward as possible.
Step 102, object candidate window is input in four combination face characteristic graders in parallel and carried out further
Screening, obtain face candidate window.
Combination face characteristic be based between face characteristic there are the characteristics of relatively-stationary spatial relation, to face
Single features carry out the result of combination of two.Specifically, the object candidate window obtained from step 101 is input to four and is based on
It is screened in the combination face characteristic grader of depth convolutional network, after obtaining face candidate window, is pressed down using non-maximum value
Algorithm processed carries out window duplicate removal processing to face candidate window.
It is illustrated in figure 3 the depth convolutional network structure chart of an assemblage characteristic grader in the specific embodiment of the invention;
As can be seen from Figure 3, which is made of 7 convolutional layers and 3 pond layers, to promote the feature that more horn of plenty is abstracted, and
Its specific training process is divided into two stages, and first stage is the pre-training of model, and second stage is then combination face
The training of property data base, to promote whole detection performance.
Step 103, face candidate window is input in a multitask depth convolutional network, carries out further face
Detection and window correction, and the face to detecting carries out angle classification.
Specifically, it in embodiments of the present invention, will be inputted by the face candidate window after non-maximum value restrainable algorithms
Into a multitask depth convolutional network, one side face candidate window can further be screened and window correction, another
Aspect can carry out facial angle classification to the window that screening passes through, to judge whether the candidate window is non-face window of listening to the teacher.
It is illustrated in figure 4 the structure chart of a multitask depth convolutional network in the specific embodiment of the invention;It can from Fig. 4
Know, which is made of 5 convolutional layers and 3 pond layers, and output layer is made of three parts, respectively carrying out face
Detection, face window return and facial angle classification, after the multitask depth convolutional network, can obtain detection people
Listen to the teacher number and non-number of listening to the teacher in face, the specific training process of the network are also classified into two stages:First stage is
The fine tuning training of model, second stage is then the training for combining facial feature database, to promote whole detection performance.
Step 104, it is listened to the teacher the detected value of rate according to what facial angle was classified as a result, obtaining classroom in picture to be detected.
In embodiments of the present invention, by the facial angle classification results and Face datection that are obtained in step 103 as a result, will
As classroom total number of persons, the former described the latter listen to the teacher total number of persons as classroom, and the ratio of the two is then listened to the teacher for specific time period classroom
The detected value of rate.
Preferably, before step 104, following steps are further included:
By picture to be detected be input to a multitask depth convolutional network as a result, obtain picture in total number of persons estimation
Value, and using the value as the estimated value of a certain moment classroom total number of persons.
It is illustrated in figure 5 the schematic diagram of the multitask depth convolutional network structure;As can be seen from Figure 5, the network structure is by 3
A convolutional layer and 3 pond layers compositions, and output layer is made of two parts, respectively carrying out the acquisition of crowd density response diagram
With the estimation of total number of persons, but crowd density response diagram is only intended to the training of auxiliary demographics, by the cross-training of the two,
The accuracy of entire model Population size estimation can be promoted.And its specific training process is also classified into two stages:First stage
For the pre-training of model, second stage is then the fine tuning training of model, with the total evaluation performance of lift scheme.
Fig. 6 is a kind of system architecture diagram of face detection system of the present invention.As shown in fig. 6, a kind of Face datection of the present invention
System, including:
Object candidate window generation unit 601, for using classroom picture to be detected as input, utilizing selective search
Method generates object candidate window.Here selective search method can be grouped according to region subdivision, Biodiversity Characteristics segmentation with
And location information reduces the production of invalid window on a large scale with reference to the candidate object window come in quick obtaining picture to be detected
It is raw;
Face candidate window acquiring unit 602, it is special for object candidate window to be input to four combination faces in parallel
It is further screened in sign grader, obtains face candidate window.In the present invention, combination face characteristic is based on face
There are the characteristics of relatively-stationary spatial relation between feature, face single features are carried out with the result of combination of two.People
Four assemblage characteristic graders of face candidate window acquiring unit 502 share identical depth convolutional network structure, the network
Structure is made of 7 convolutional layers and 3 pond layers;
Facial angle taxon 603, for face candidate window to be input in a multitask depth convolutional network,
Further Face datection and window correction are carried out, and the face to detecting carries out angle classification.Here multitask depth
Convolutional network is made of three tasks, and respectively Face datection, window return and facial angle is classified, and passes through multitask depth
After spending convolutional network, listen to the teacher number and non-number of listening to the teacher in detection face can be obtained, multitask depth convolutional network is by 5
Convolutional layer and 3 pond layer compositions.
Rate of listening to the teacher detected value computing unit 604 is taught for what is classified according to facial angle as a result, obtaining in picture to be detected
Room is listened to the teacher the detected value of rate.Specifically, according to facial angle classification results and Face datection as a result, using described the latter as classroom
Total number of persons, the former listens to the teacher total number of persons as classroom, and the ratio of the two is then listened to the teacher the detected value of rate for specific time period classroom.
Preferably, the face detection system of the present invention further includes:It listens to the teacher total number of persons estimation unit, for by picture to be detected
Be input to a multitask depth convolutional network as a result, obtain the estimated value of total number of persons in picture, and using the value as certain for the moment
Carve the estimated value of classroom total number of persons.
Fig. 7 is a kind of step flow chart of the rate appraisal procedure of listening to the teacher based on Face datection of the present invention.As shown in fig. 7, this
A kind of rate appraisal procedure of listening to the teacher based on Face datection is invented, is included the following steps:
Step 701, the rate detected value of listening to the teacher obtained according to method for detecting human face, according to a predetermined threshold value to the detected value
It is screened, deletes improper data.
Specifically, in present example, using the rate detected value of listening to the teacher obtained from method for detecting human face according to pre-
If threshold value screens detected value, undesirable detection data is deleted.
Preferably, the predetermined threshold value is generally designated as 1/2 of the maximum value of all detected values in detected video, described
Undesirable detection data is then appointed as the detected value less than the threshold value, after deleting undesirable detection data,
The rate detected value of listening to the teacher of video to be detected can be obtained, in this, as subsequent calibration basis.
Step 702, the inspection of listen to the teacher total number of persons estimated value and method for detecting human face acquisition obtained according to method for detecting human face
Measured value obtains the number not detected in classroom.
Specifically, in present example, the acquisition methods for not detecting number are as follows, according to the to be detected of input
Video, carries out its video frame Face datection respectively and demographics obtain corresponding detected value, by the former Face datection number
As detection number, the latter as total number of persons, the two subtract each other to obtain described in do not detect number.
Step 703, the rate detected value of listening to the teacher obtained according to above-mentioned method for detecting human face, obtains its minimum as step
The rate of listening to the teacher of number is not detected described in 702, the calibration for rate of listening to the teacher with completion.
Specifically, in the specific embodiment of the invention, the minimum detected value is by the detected value that is obtained in step 701
Minimum value do not detect number as minimum detected value, and using what is obtained in step 702, calculate and obtain in non-detection number
It listens to the teacher number, to complete final rate calibration of listening to the teacher.The final rate of listening to the teacher number and does not detect people by listening to the teacher for detected value
The ratio acquisition of several listen to the teacher the sum of number and total numbers of persons of demographics.
Fig. 8 is a kind of system architecture diagram of the rate assessment system of listening to the teacher based on Face datection of the present invention.As shown in figure 8, this
A kind of rate assessment system of listening to the teacher based on Face datection is invented, including:
Detected value screening unit 801, for the rate detected value of listening to the teacher obtained according to method for detecting human face, according to a default threshold
Value screens the detected value, deletes improper data.
Specifically, in present example, using the rate detected value of listening to the teacher obtained from method for detecting human face according to pre-
If threshold value screens detected value, undesirable detection data is deleted.Preferably, the predetermined threshold value is generally designated as
The 1/2 of the maximum value of all detected values in detected video, the undesirable detection data are then appointed as being less than described
The detected value of threshold value after deleting undesirable detection data, can obtain the rate detected value of listening to the teacher of video to be detected, with this
As subsequent calibration basis.
Number acquiring unit 802 is not detected, for listen to the teacher total number of persons estimated value and the people obtained according to method for detecting human face
The detected value that face detecting method obtains obtains the number not detected in classroom.
Specifically, in the specific embodiment of the invention, the acquisition methods for not detecting number are as follows, according to input
Video to be detected, carries out its video frame Face datection respectively and demographics obtain corresponding detected value, by the former face
Testing number is as detection number, and for the latter as total number of persons, the two, which subtracts each other to obtain, described does not detect number.
Alignment unit 803 for the rate detected value of listening to the teacher obtained according to above-mentioned method for detecting human face, obtains its minimum work
The rate of listening to the teacher of number is not detected described in number acquiring unit 802 not detect, the calibration for rate of listening to the teacher with completion.
Specifically, in the specific embodiment of the invention, the minimum detected value is by the minimum value in the detected value that obtains
As minimum detected value, and using do not detect obtained in number acquiring unit 702 do not detect number, calculate to obtain and do not detect people
Number of listening to the teacher in number, to complete final rate calibration of listening to the teacher.Final the listen to the teacher number and not of the rate by detected value of listening to the teacher
The ratio acquisition of listen to the teacher the sum of number and the total number of persons of demographics of detection number.
In conclusion a kind of method for detecting human face of the present invention and rate appraisal procedure of listening to the teacher based on Face datection, system according to
According to demand using picture to be detected or video as input, Face datection and face angle are carried out to input using method for detecting human face
Degree classification is obtained the detected value for rate of listening to the teacher, is then carried out the estimation of total number of persons, final basis to input using people counting algorithm
Classroom is listened to the teacher the detected value of rate and the estimated value of total number of persons, and final rate assessment of listening to the teacher is obtained using rate calibration method is listened to the teacher
Value, inventive energy is good, accuracy rate is high, easy realization, is conducive to the quantizating index for teaching system being commented to provide the rate of listening to the teacher for the modern times.
The above-described embodiments merely illustrate the principles and effects of the present invention, and is not intended to limit the present invention.Any
Field technology personnel can modify above-described embodiment and changed under the spirit and scope without prejudice to the present invention.Therefore,
The scope of the present invention, should be as listed by claims.
Claims (10)
1. a kind of method for detecting human face, includes the following steps:
Using classroom picture to be detected as input, object candidate window is generated using selective search method for step 1;
Object candidate window is input in the combination face characteristic grader of multiple parallel connections and is further sieved by step 2
Choosing obtains face candidate window;
Face candidate window is input in a multitask depth convolutional network, carries out further Face datection by step 3
And window correction, and the face to detecting carries out angle classification
Step 4 is listened to the teacher the detected value of rate according to what facial angle was classified as a result, obtaining classroom in picture to be detected.
2. a kind of method for detecting human face as described in claim 1, which is characterized in that step 1 further comprises:
The target object of acquisition arbitrary dimension is removed using region subdivision grouping algorithm;
All situations that may be present are handled using diversified feature policy;
Combining position information, the region obtained by region subdivision grouping algorithm and diversified Image Segmentation Methods Based on Features strategy, is selectively searched
Suo Caiyong location informations are ranked up screening come the region to generation, filter out the region of sequence rearward.
3. a kind of method for detecting human face as described in claim 1, it is characterised in that:In step 2, obtained from step 1
Object candidate window is input in four combination face characteristic graders based on depth convolutional network and is screened, and obtains face
After candidate window, window duplicate removal processing is carried out to face candidate window using non-maximum value restrainable algorithms.
4. a kind of method for detecting human face as claimed in claim 3, it is characterised in that:Four assemblage characteristic graders share
Identical depth convolutional network structure, the network structure include 7 convolutional layers and 3 pond layers.
5. a kind of method for detecting human face as claimed in claim 4, it is characterised in that:In step 4, obtained according to step 3
Facial angle classification results and Face datection as a result, using described the latter as classroom total number of persons, the former listens to the teacher always as classroom
Number, the ratio of the two are then listened to the teacher the detected value of rate for specific time period classroom.
6. a kind of method for detecting human face as described in claim 1, which is characterized in that before step 4, further include:
By picture to be detected be input to a multitask depth convolutional network as a result, obtain picture in total number of persons estimated value, and
Using the value as the estimated value of a certain moment classroom total number of persons.
7. a kind of method for detecting human face as claimed in claim 6, which is characterized in that the multitask depth convolutional network includes
3 convolutional layers and 2 pond layers, multitask include crowd density response diagram and Population size estimation.
8. a kind of face detection system, including:
Object candidate window generation unit, for using classroom picture to be detected as input, being given birth to using selective search method
Into object candidate window;
Face candidate window acquiring unit, for object candidate window to be input to the combination face characteristic grader of multiple parallel connections
It is middle further to be screened, obtain face candidate window;
Facial angle taxon, for face candidate window to be input in a multitask depth convolutional network, into traveling
The Face datection and window correction of one step, and the face to detecting carries out angle classification
Rate of listening to the teacher detected value computing unit is listened to the teacher for what is classified according to facial angle as a result, obtaining classroom in picture to be detected
The detected value of rate.
9. a kind of rate appraisal procedure of listening to the teacher based on Face datection, includes the following steps:
Step 1 according to the rate detected value of listening to the teacher that method for detecting human face obtains, carries out the detected value according to a predetermined threshold value
Improper data is deleted in screening;
Step 2, the detected value of listen to the teacher total number of persons estimated value and method for detecting human face acquisition obtained according to method for detecting human face,
Obtain the number not detected in classroom;
Step 3 according to the rate detected value of listening to the teacher that above-mentioned method for detecting human face obtains, obtains its minimum as described in step 2
The rate of listening to the teacher of number is not detected, the calibration for rate of listening to the teacher with completion.
10. a kind of rate assessment system of listening to the teacher based on Face datection, including:
Detected value screening unit, for the rate detected value of listening to the teacher obtained according to method for detecting human face, according to a predetermined threshold value to institute
It states detected value to be screened, deletes improper data;
Number acquiring unit is not detected, for listen to the teacher total number of persons estimated value and the Face datection side obtained according to method for detecting human face
The detected value that method obtains obtains the number not detected in classroom;
Alignment unit for the rate detected value of listening to the teacher obtained according to above-mentioned method for detecting human face, obtains the conduct of its minimum and does not examine
The rate of listening to the teacher of number is not detected described in survey number acquiring unit, the calibration for rate of listening to the teacher with completion.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810073494.4A CN108256490A (en) | 2018-01-25 | 2018-01-25 | Method for detecting human face and listen to the teacher rate appraisal procedure, system based on Face datection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810073494.4A CN108256490A (en) | 2018-01-25 | 2018-01-25 | Method for detecting human face and listen to the teacher rate appraisal procedure, system based on Face datection |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108256490A true CN108256490A (en) | 2018-07-06 |
Family
ID=62742537
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810073494.4A Pending CN108256490A (en) | 2018-01-25 | 2018-01-25 | Method for detecting human face and listen to the teacher rate appraisal procedure, system based on Face datection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108256490A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109766823A (en) * | 2019-01-07 | 2019-05-17 | 浙江大学 | A kind of high-definition remote sensing ship detecting method based on deep layer convolutional neural networks |
CN110796029A (en) * | 2019-10-11 | 2020-02-14 | 北京达佳互联信息技术有限公司 | Face correction and model training method and device, electronic equipment and storage medium |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104732601A (en) * | 2014-11-19 | 2015-06-24 | 东北大学 | Automatic high-recognition-rate attendance checking device and method based on face recognition technology |
CN105825189A (en) * | 2016-03-21 | 2016-08-03 | 浙江工商大学 | Device for automatically analyzing attendance rate and class concentration degree of college students |
CN106228293A (en) * | 2016-07-18 | 2016-12-14 | 重庆中科云丛科技有限公司 | teaching evaluation method and system |
CN107169900A (en) * | 2017-05-19 | 2017-09-15 | 南京信息工程大学 | A kind of student listens to the teacher rate detection method |
-
2018
- 2018-01-25 CN CN201810073494.4A patent/CN108256490A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN104732601A (en) * | 2014-11-19 | 2015-06-24 | 东北大学 | Automatic high-recognition-rate attendance checking device and method based on face recognition technology |
CN105825189A (en) * | 2016-03-21 | 2016-08-03 | 浙江工商大学 | Device for automatically analyzing attendance rate and class concentration degree of college students |
CN106228293A (en) * | 2016-07-18 | 2016-12-14 | 重庆中科云丛科技有限公司 | teaching evaluation method and system |
CN107169900A (en) * | 2017-05-19 | 2017-09-15 | 南京信息工程大学 | A kind of student listens to the teacher rate detection method |
Non-Patent Citations (1)
Title |
---|
XIANBO YU ET AL.: "Face Detection:A Deep Convolutional Network Method Based on Grouped Facial Part", 《2017 IEEE 2ND ADVANCED INFORMATION TECHNOLOGY, ELECTRONIC AND AUTOMATION CONTROL CONFERENCE (IAEAC)》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109766823A (en) * | 2019-01-07 | 2019-05-17 | 浙江大学 | A kind of high-definition remote sensing ship detecting method based on deep layer convolutional neural networks |
CN110796029A (en) * | 2019-10-11 | 2020-02-14 | 北京达佳互联信息技术有限公司 | Face correction and model training method and device, electronic equipment and storage medium |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109801256B (en) | Image aesthetic quality assessment method based on region of interest and global features | |
CN107945153A (en) | A kind of road surface crack detection method based on deep learning | |
CN109697434A (en) | A kind of Activity recognition method, apparatus and storage medium | |
CN108073888A (en) | A kind of teaching auxiliary and the teaching auxiliary system using this method | |
CN108510004B (en) | Cell classification method and system based on deep residual error network | |
CN105160678A (en) | Convolutional-neural-network-based reference-free three-dimensional image quality evaluation method | |
CN108416314B (en) | Picture important face detection method | |
CN109508671A (en) | A kind of video accident detection system and method based on Weakly supervised study | |
CN107871101A (en) | A kind of method for detecting human face and device | |
CN109034036A (en) | A kind of video analysis method, Method of Teaching Quality Evaluation and system, computer readable storage medium | |
CN103164687B (en) | A kind of method and system of pornographic image detecting | |
CN104537647A (en) | Target detection method and device | |
CN103971106B (en) | Various visual angles facial image gender identification method and device | |
CN106407369A (en) | Photo management method and system based on deep learning face recognition | |
CN108305253A (en) | A kind of pathology full slice diagnostic method based on more multiplying power deep learnings | |
CN109544537A (en) | The fast automatic analysis method of hip joint x-ray image | |
CN111145145B (en) | Image surface defect detection method based on MobileNet | |
CN110009628A (en) | A kind of automatic testing method for polymorphic target in continuous two dimensional image | |
CN110163567A (en) | Classroom roll calling system based on multitask concatenated convolutional neural network | |
CN110349167A (en) | A kind of image instance dividing method and device | |
CN107480607A (en) | A kind of method that standing Face datection positions in intelligent recording and broadcasting system | |
Du et al. | Multi-focus image fusion algorithm based on pulse coupled neural networks and modified decision map | |
CN112364754A (en) | Bolt defect detection method and system | |
CN109523558A (en) | A kind of portrait dividing method and system | |
CN108256490A (en) | Method for detecting human face and listen to the teacher rate appraisal procedure, system based on Face datection |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20180706 |
|
WD01 | Invention patent application deemed withdrawn after publication |