Transferability of a Convolutional Neural Network (CNN) to Measure Traffic Density
Whereas detecting individual vehicles in a video image using a convolutional neural network (CNN) prevails for traffic surveillance, CNNs also have been successfully adapted to counting vehicles via a regression method, which conveys the advantages of simplifying the model structure, and inference t...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2021-05-01
|
| Series: | Electronics |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2079-9292/10/10/1189 |
| _version_ | 1797533858914107392 |
|---|---|
| author | Jiyong Chung Gyeongjun Kim Keemin Sohn |
| author_facet | Jiyong Chung Gyeongjun Kim Keemin Sohn |
| author_sort | Jiyong Chung |
| collection | DOAJ |
| description | Whereas detecting individual vehicles in a video image using a convolutional neural network (CNN) prevails for traffic surveillance, CNNs also have been successfully adapted to counting vehicles via a regression method, which conveys the advantages of simplifying the model structure, and inference time can be reduced in the field. This model also demands much less human effort to tag images with labels. The number of vehicles in an image becomes the label, rather than bounding boxes drawn around every single vehicle. Nonetheless, the labeling task takes considerable time whenever a CNN model is trained and tested for a new road segment. There are two ways to alleviate the human effort involved in using this method. A previous study used a pseudo label pre-training method, and another study employed an image synthesis method to solve the problem. Besides these two methods, we investigated the model transferability to reduce the labeling effort. Using a CNN that was fully trained on images of a road segment, we devised a robust way to utilize the trained model for another site by transforming the model output with a simple quadratic equation. The utility of the proposed method was confirmed at the expense of a minute amount of deterioration in accuracy. |
| first_indexed | 2024-03-10T11:21:37Z |
| format | Article |
| id | doaj.art-15e429256fe5486d9de513b87b69651a |
| institution | Directory Open Access Journal |
| issn | 2079-9292 |
| language | English |
| last_indexed | 2024-03-10T11:21:37Z |
| publishDate | 2021-05-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Electronics |
| spelling | doaj.art-15e429256fe5486d9de513b87b69651a2023-11-21T19:57:05ZengMDPI AGElectronics2079-92922021-05-011010118910.3390/electronics10101189Transferability of a Convolutional Neural Network (CNN) to Measure Traffic DensityJiyong Chung0Gyeongjun Kim1Keemin Sohn2Laboratory of Big-Data Applications in Public Sectors, Department of Urban Engineering, Chung-Ang University, 84 Heukseok-ro, Dongjak-gu, Seoul 156-756, KoreaLaboratory of Big-Data Applications in Public Sectors, Department of Urban Engineering, Chung-Ang University, 84 Heukseok-ro, Dongjak-gu, Seoul 156-756, KoreaLaboratory of Big-Data Applications in Public Sectors, Department of Urban Engineering, Chung-Ang University, 84 Heukseok-ro, Dongjak-gu, Seoul 156-756, KoreaWhereas detecting individual vehicles in a video image using a convolutional neural network (CNN) prevails for traffic surveillance, CNNs also have been successfully adapted to counting vehicles via a regression method, which conveys the advantages of simplifying the model structure, and inference time can be reduced in the field. This model also demands much less human effort to tag images with labels. The number of vehicles in an image becomes the label, rather than bounding boxes drawn around every single vehicle. Nonetheless, the labeling task takes considerable time whenever a CNN model is trained and tested for a new road segment. There are two ways to alleviate the human effort involved in using this method. A previous study used a pseudo label pre-training method, and another study employed an image synthesis method to solve the problem. Besides these two methods, we investigated the model transferability to reduce the labeling effort. Using a CNN that was fully trained on images of a road segment, we devised a robust way to utilize the trained model for another site by transforming the model output with a simple quadratic equation. The utility of the proposed method was confirmed at the expense of a minute amount of deterioration in accuracy.https://www.mdpi.com/2079-9292/10/10/1189model transferabilitytraffic surveillanceconvolutional neural networktraffic density |
| spellingShingle | Jiyong Chung Gyeongjun Kim Keemin Sohn Transferability of a Convolutional Neural Network (CNN) to Measure Traffic Density Electronics model transferability traffic surveillance convolutional neural network traffic density |
| title | Transferability of a Convolutional Neural Network (CNN) to Measure Traffic Density |
| title_full | Transferability of a Convolutional Neural Network (CNN) to Measure Traffic Density |
| title_fullStr | Transferability of a Convolutional Neural Network (CNN) to Measure Traffic Density |
| title_full_unstemmed | Transferability of a Convolutional Neural Network (CNN) to Measure Traffic Density |
| title_short | Transferability of a Convolutional Neural Network (CNN) to Measure Traffic Density |
| title_sort | transferability of a convolutional neural network cnn to measure traffic density |
| topic | model transferability traffic surveillance convolutional neural network traffic density |
| url | https://www.mdpi.com/2079-9292/10/10/1189 |
| work_keys_str_mv | AT jiyongchung transferabilityofaconvolutionalneuralnetworkcnntomeasuretrafficdensity AT gyeongjunkim transferabilityofaconvolutionalneuralnetworkcnntomeasuretrafficdensity AT keeminsohn transferabilityofaconvolutionalneuralnetworkcnntomeasuretrafficdensity |