CN112488167A - Rubbish identification and classification system based on improved EfficientNet network - Google Patents
Rubbish identification and classification system based on improved EfficientNet network Download PDFInfo
- Publication number
- CN112488167A CN112488167A CN202011303297.0A CN202011303297A CN112488167A CN 112488167 A CN112488167 A CN 112488167A CN 202011303297 A CN202011303297 A CN 202011303297A CN 112488167 A CN112488167 A CN 112488167A
- Authority
- CN
- China
- Prior art keywords
- improved
- garbage
- learning rate
- network
- efficientnet network
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/77—Processing image or video features in feature spaces; using data integration or data reduction, e.g. principal component analysis [PCA] or independent component analysis [ICA] or self-organising maps [SOM]; Blind source separation
- G06V10/774—Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
- G06V2201/06—Recognition of objects for industrial automation
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Software Systems (AREA)
- Computing Systems (AREA)
- General Health & Medical Sciences (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Mathematical Physics (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Image Analysis (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention aims to solve the technical problem of providing a garbage classification system based on an improved EfficientNet network, and the garbage classification is more accurate and the classification speed is higher by utilizing an improved EfficientNet network classification algorithm. The method comprises the following specific steps: (1) collecting various garbage through picture collecting equipment, and establishing a database; (2) the method comprises the steps of preprocessing the collected garbage pictures, wherein common preprocessing methods comprise gray processing, median filtering processing and image enhancement, and are used for enhancing interesting features in the images and inhibiting uninteresting features in the images, so that the quality of the images can be effectively improved. Secondly, dividing the preprocessed data; (3) adopting a migration learning method, keeping the original weight of the EfficientNet network on the ImageNet data set, training a new weight through fine tuning, and storing the trained EfficientNet network model; (4) and carrying out classification and identification on the preprocessed data set through the improved EfficientNet network model to obtain a classification result.
Description
Technical Field
The invention relates to the field of target detection, in particular to a garbage identification and classification system based on an improved EfficientNet network
Background
With the economic development, the living standard of the people is obviously improved, but the various domestic wastes are increased at an incredible speed. If a large amount of domestic waste is not handled in time, not only city face is influenced, also can harm people's health. Today, a saving and environment-friendly society is advocated, not only increasing amounts of garbage need to be treated, but also available resources in the garbage need to be recovered. Garbage classification is an efficient resource recovery method and a social problem concerning the sustainable development of the livelihood and society
The problem of garbage classification has become a focus at present, and more people study the garbage classification through deep learning, wherein an artificial neural network makes a great contribution in the field of garbage classification. The predecessors used the BP neural network to classify garbage, the improved BP to classify garbage, and the convolutional neural network to classify garbage. These still present some problems:
(1) the effect of garbage identification and classification is poor. Various neural networks are used to classify garbage, and the application of these neural networks to garbage classification has been shown to be excellent. However, the final garbage classification effect is not good enough, and the classification precision of some classification systems is very low.
(2) The garbage classification algorithm is slow in identification speed. Due to the large variation of the shape and color of the garbage, it is not easy to manually extract the category features, and the data volume is large. Therefore, the speed of the classification algorithm for garbage classification and identification is very low, and the real-time requirement of the embedded equipment in the production field cannot be met.
Therefore, it is necessary to develop a garbage classification system having a high classification effect and a fast operation speed.
Disclosure of Invention
The invention aims to solve the technical problem of providing a garbage identification and classification system based on an improved EfficientNet network, and realizing more accurate garbage classification and higher classification speed by utilizing an improved EfficientNet network classification algorithm. In order to solve the technical problems, the invention adopts the technical scheme that: the garbage identification and classification system based on the improved EfficientNet network specifically comprises the following steps, and a related flow chart is shown in a figure 1:
(1) collecting various garbage through picture collecting equipment, and establishing a database;
(2) preprocessing the acquired images of the garbage, preprocessing the acquired images of the garbage by adopting methods such as gray processing, median filtering processing, image enhancement and the like, and then dividing a preprocessed data set;
(3) adopting a migration learning method, keeping the original weight of the EfficientNet network on the ImageNet data set, training a new weight through fine tuning, and storing the trained EfficientNet network model;
(4) classifying and identifying the preprocessed data set by using the improved EfficientNet network model to obtain a classification identification result;
the garbage identification and classification system based on the improved EfficientNet network, provided by the invention, has the advantages that the algorithm of the EfficientNet is optimized; specifically, the learning rate of the whole algorithm is adjusted by using a cosine learning rate attenuation mode with warm-up (preheating), and common modes for changing the learning rate include an exponential attenuation mode, a gradual attenuation mode and a polynomial attenuation mode. Since the weights (weights) of the model are initialized randomly when the training is started, a larger learning rate is selected at the moment, and the model may be unstable, so that a smaller learning rate is used when the training is started to preheat the learning rate, and the preset learning rate is modified to train when the model is stable, so that the convergence rate of the model is higher, and the model effect is better. And secondly, the learning rate is reduced by utilizing a cosine annealing mode, so that the precision is improved. The combination of the two modes improves the recognition rate and the speed of the system.
The improved specific process comprises the following steps:
the improvement is mainly to change the calculation mode of the learning rate, and the new learning rate calculation method is mainly divided into two stages, namely a Warmup stage and a cosine annealing stage.
The Warmup stage:
(1) firstly, selecting a smaller preheating learning rate, training a network model, training epochs or steps (such as 4 epochs, 10000steps), then increasing each step by a little, and observing a training result;
(2) when the trained model is stable, increasing the learning rate to a preset learning rate;
cosine annealing stage: gradually reducing the learning rate from an initial value according to a cosine function after the learning rate of the Warmup stage, and observing and training
Training results;
after the two phases are combined, the specific learning rate is calculated as follows:
(1) firstly, initializing a preheating initial learning rate, a preheating basic learning rate, epochs (total iteration times), a Warmup _ epoch (preheating iteration times), a batch _ size, a sample number and a reserved step number;
(2) calculating the learning rate of the Warmup according to a formula;
(3) after the Warmup is finished, calculating the cosine annealing learning rate according to a formula;
(4) the cosine annealing learning rate is further calculated and judged, if the reserved step number is larger than 0, whether the current step number is larger than the preheating step number plus the reserved step number or not is judged, if yes, the calculated learning rate is returned, and if not, the basic learning rate after Warmup is used;
(5) when the preheating step number is more than 0, if the current step number is less than the preheating step number, returning the learning rate of the Warmup at that time, otherwise, directly returning to the cosine annealing calculation of the second step;
(6) if the last current step number is larger than the total step number, the learning rate is 0, otherwise, the current calculated learning rate is returned;
drawings
The following further detailed description of embodiments of the invention is made with reference to the accompanying drawings:
FIG. 1 is a schematic flow chart of a garbage identification and classification system based on an improved EfficientNet network
FIG. 2 is a flow chart of an improved EfficientNet network algorithm
FIG. 3 is a schematic diagram of cosine annealing with Warmup
FIG. 4 is a plot of cosine learning rate decay versus stepwise learning rate decay.
Claims (5)
1. A garbage identification and classification system based on an improved EfficientNet network is characterized by comprising the following steps:
(1) collecting various garbage through picture collecting equipment, and establishing a database;
(2) preprocessing the acquired images of the garbage, preprocessing the acquired images of the garbage by adopting methods such as gray processing, median filtering processing, image enhancement and the like, and then dividing a preprocessed data set;
(3) adopting a migration learning method, keeping the original weight of the EfficientNet network on the ImageNet data set, training a new weight through fine tuning, and storing the trained EfficientNet network model;
(4) and classifying and identifying the preprocessed data set by using the improved EfficientNet network model to obtain a classification and identification result.
2. The improved EfficientNet network-based garbage recognition and classification system according to claim 1, wherein in the step (1), a raw image of garbage is collected by a hardware device such as a camera, a video camera, a mobile phone, etc., and the pictures are collected to construct a picture database.
3. The improved EfficientNet network-based garbage identification and classification system according to claim 1, wherein in the step (2), the original database of the picture is preprocessed, and unnecessary information such as noise in the picture is removed by median filtering and a picture enhancement method, wherein the picture enhancement method generally adopts a frequency domain method and a spatial domain method.
4. The improved EfficientNet network-based garbage recognition and classification system of claim 1, wherein in the step (3), the EfficientNet network trained on the ImageNet data set is transplanted by using a migration learning method, and then the network is retrained by using the preprocessed pictures, so that the weight of the network is modified again to meet the garbage classification condition.
5. The improved EfficientNet network-based garbage recognition and classification system according to claim 1, wherein in the step (4), the EfficientNet network is improved based on the transfer learning, and the garbage is classified by using the improved EfficientNet network. The improved content is mainly to improve the learning rate of the network, and the learning rate is modified in a mode of cosine learning rate attenuation with Warmup, so that the identification precision and the running speed of the EfficientNet network model are improved. Specifically, the learning rate of the whole algorithm is adjusted by using a cosine annealing learning rate scheduling method with warm-up (preheating). Since the weights (weights) of the model are initialized randomly when the training is started, a larger learning rate is selected at the moment, and the model may be unstable, so that a smaller learning rate is used when the training is started to preheat the learning rate, and the preset learning rate is modified to train when the model is stable, so that the convergence rate of the model is higher, and the model effect is better. And secondly, the learning rate is reduced by utilizing a cosine annealing mode, so that the precision is improved. The combination of the two modes improves the recognition rate and the speed of the system.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011303297.0A CN112488167A (en) | 2020-11-19 | 2020-11-19 | Rubbish identification and classification system based on improved EfficientNet network |
| LU500063A LU500063B1 (en) | 2020-11-19 | 2021-04-20 | Garbage identification and classification system based on improved efficient net |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011303297.0A CN112488167A (en) | 2020-11-19 | 2020-11-19 | Rubbish identification and classification system based on improved EfficientNet network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112488167A true CN112488167A (en) | 2021-03-12 |
Family
ID=74932046
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011303297.0A Pending CN112488167A (en) | 2020-11-19 | 2020-11-19 | Rubbish identification and classification system based on improved EfficientNet network |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN112488167A (en) |
| LU (1) | LU500063B1 (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113610163A (en) * | 2021-08-09 | 2021-11-05 | 安徽工业大学 | Knowledge distillation-based lightweight apple leaf disease identification method |
| CN113657143A (en) * | 2021-06-25 | 2021-11-16 | 中国计量大学 | Garbage classification method based on classification and detection joint judgment |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110786824A (en) * | 2019-12-02 | 2020-02-14 | 中山大学 | Coarse marking fundus oculi illumination bleeding lesion detection method and system based on bounding box correction network |
| CN110852420A (en) * | 2019-11-11 | 2020-02-28 | 北京智能工场科技有限公司 | Garbage classification method based on artificial intelligence |
| CN110884791A (en) * | 2019-11-28 | 2020-03-17 | 石家庄邮电职业技术学院(中国邮政集团公司培训中心) | Vision garbage classification system and classification method based on TensorFlow |
| CN111079639A (en) * | 2019-12-13 | 2020-04-28 | 中国平安财产保险股份有限公司 | Method, device and equipment for constructing garbage image classification model and storage medium |
| CN111611949A (en) * | 2020-05-25 | 2020-09-01 | 陈金山 | Intelligent urban water area garbage density detection method based on CIM and block chain |
| CN111723772A (en) * | 2020-06-30 | 2020-09-29 | 平安国际智慧城市科技股份有限公司 | Perishable garbage identification method and device based on image identification and computer equipment |
| CN111768406A (en) * | 2020-08-05 | 2020-10-13 | 赛维森(广州)医疗科技服务有限公司 | Cell image processing method, device, equipment and storage medium |
| CN111783841A (en) * | 2020-06-09 | 2020-10-16 | 中科院成都信息技术股份有限公司 | Garbage classification method, system and medium based on transfer learning and model fusion |
-
2020
- 2020-11-19 CN CN202011303297.0A patent/CN112488167A/en active Pending
-
2021
- 2021-04-20 LU LU500063A patent/LU500063B1/en active IP Right Grant
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110852420A (en) * | 2019-11-11 | 2020-02-28 | 北京智能工场科技有限公司 | Garbage classification method based on artificial intelligence |
| CN110884791A (en) * | 2019-11-28 | 2020-03-17 | 石家庄邮电职业技术学院(中国邮政集团公司培训中心) | Vision garbage classification system and classification method based on TensorFlow |
| CN110786824A (en) * | 2019-12-02 | 2020-02-14 | 中山大学 | Coarse marking fundus oculi illumination bleeding lesion detection method and system based on bounding box correction network |
| CN111079639A (en) * | 2019-12-13 | 2020-04-28 | 中国平安财产保险股份有限公司 | Method, device and equipment for constructing garbage image classification model and storage medium |
| CN111611949A (en) * | 2020-05-25 | 2020-09-01 | 陈金山 | Intelligent urban water area garbage density detection method based on CIM and block chain |
| CN111783841A (en) * | 2020-06-09 | 2020-10-16 | 中科院成都信息技术股份有限公司 | Garbage classification method, system and medium based on transfer learning and model fusion |
| CN111723772A (en) * | 2020-06-30 | 2020-09-29 | 平安国际智慧城市科技股份有限公司 | Perishable garbage identification method and device based on image identification and computer equipment |
| CN111768406A (en) * | 2020-08-05 | 2020-10-13 | 赛维森(广州)医疗科技服务有限公司 | Cell image processing method, device, equipment and storage medium |
Non-Patent Citations (2)
| Title |
|---|
| MING ZENG 等: "PublicGarbageNet : A Deep Learning Framework for Public Garbage Classification", 《PROCEEDINGS OF THE 39TH CHINESE CONTROL CONFERENCE》, 9 September 2020 (2020-09-09), pages 3 * |
| 高明 等: "基于新型空间注意力机制和迁移学习的垃圾图像分类算法", 《系统工程理论与实践》, 28 October 2020 (2020-10-28), pages 3 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113657143A (en) * | 2021-06-25 | 2021-11-16 | 中国计量大学 | Garbage classification method based on classification and detection joint judgment |
| CN113657143B (en) * | 2021-06-25 | 2023-06-23 | 中国计量大学 | Garbage classification method based on classification and detection combined judgment |
| CN113610163A (en) * | 2021-08-09 | 2021-11-05 | 安徽工业大学 | Knowledge distillation-based lightweight apple leaf disease identification method |
Also Published As
| Publication number | Publication date |
|---|---|
| LU500063B1 (en) | 2022-05-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110555368B (en) | Fall-down behavior identification method based on three-dimensional convolutional neural network | |
| CN100367294C (en) | Method for dividing human body skin area from color digital images and video graphs | |
| CN110135386B (en) | Human body action recognition method and system based on deep learning | |
| CN111008639B (en) | License plate character recognition method based on attention mechanism | |
| CN110084149B (en) | Face verification method based on hard sample quadruple dynamic boundary loss function | |
| CN111680702B (en) | Method for realizing weak supervision image significance detection by using detection frame | |
| CN110807757B (en) | Image quality evaluation method and device based on artificial intelligence and computer equipment | |
| CN111310718A (en) | High-accuracy detection and comparison method for face-shielding image | |
| CN112488167A (en) | Rubbish identification and classification system based on improved EfficientNet network | |
| CN108520212A (en) | Method for traffic sign detection based on improved convolutional neural networks | |
| CN111160481B (en) | Adas target detection method and system based on deep learning | |
| CN110472518B (en) | Fingerprint image quality judgment method based on full convolution network | |
| CN113177464B (en) | End-to-end multi-mode gait recognition method based on deep learning | |
| CN112507617A (en) | Training method of SRFlow super-resolution model and face recognition method | |
| CN113139501B (en) | Pedestrian multi-attribute identification method combining local area detection and multi-level feature grabbing | |
| CN109784368A (en) | A kind of determination method and apparatus of application program classification | |
| CN110796623A (en) | Infrared image rain removing method and device based on progressive residual error network | |
| CN111340758A (en) | Novel efficient iris image quality evaluation method based on deep neural network | |
| CN111126155B (en) | Pedestrian re-identification method for generating countermeasure network based on semantic constraint | |
| CN110633689B (en) | Face recognition model based on semi-supervised attention network | |
| CN103903017B (en) | A kind of face identification method based on adaptive soft histogram local binary patterns | |
| CN117095471B (en) | Face counterfeiting tracing method based on multi-scale characteristics | |
| CN116452469B (en) | Image defogging processing method and device based on deep learning | |
| CN113449676A (en) | Pedestrian re-identification method based on double-path mutual promotion disentanglement learning | |
| CN109472307A (en) | A kind of method and apparatus of training image disaggregated model |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210312 |