DS-DETR: A Model for Tomato Leaf Disease Segmentation and Damage Evaluation
Early blight and late blight are important factors restricting tomato yield. However, it is still a challenge to accurately and objectively detect and segment crop diseases in order to evaluate disease damage. In this paper, the Disease Segmentation Detection Transformer (DS-DETR) is proposed to seg...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2022-08-01
|
| Series: | Agronomy |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2073-4395/12/9/2023 |
| _version_ | 1827664385219231744 |
|---|---|
| author | Jianshuang Wu Changji Wen Hongrui Chen Zhenyu Ma Tian Zhang Hengqiang Su Ce Yang |
| author_facet | Jianshuang Wu Changji Wen Hongrui Chen Zhenyu Ma Tian Zhang Hengqiang Su Ce Yang |
| author_sort | Jianshuang Wu |
| collection | DOAJ |
| description | Early blight and late blight are important factors restricting tomato yield. However, it is still a challenge to accurately and objectively detect and segment crop diseases in order to evaluate disease damage. In this paper, the Disease Segmentation Detection Transformer (DS-DETR) is proposed to segment leaf disease spots efficiently based on several improvements to DETR. Additionally, a damage assessment is carried out by the area ratio of the segmented leaves to the disease spots. First, an unsupervised pre-training method was introduced into DETR with the Plant Disease Classification Dataset (PDCD) to solve the problem of the long training epochs and slow convergence speed of DETR. This method can train the Transformer structures in advance to obtain leaf disease features. Loading the pre-training model weight in DS-DETR can speed up the convergence speed of the model. Then, Spatially Modulated Co-Attention (SMCA) was used to assign Gaussian-like spatial weights to the query box of DS-DETR. The different positions in the image are trained using the query boxes with different weights to improve the accuracy of the model. Finally, an improved relative position code was added to the Transformer structure of DS-DETR. Relative position coding promotes the capture of the sequence order of input tokens by the Transformer. The spatial location feature is strengthened by establishing the location relationship between different instances. Based on these improvements, the DS-DETR model was tested on the Tomato leaf Disease Segmentation Dataset (TDSD) constructed by us. The experimental results show that the DS-DETR proposed by us achieved 0.6823 for AP<sub>mask</sub>, which improved by 12.87%, 8.25%, 3.67%, 1.95%, 10.27%, and 9.52% compared with the state-of-the-art: Mask RCNN, BlendMask, CondInst, SOLOv2, ISTR, and DETR, respectively. In addition, the disease grading accuracy reached 0.9640 according to the segmentation results given by our proposed model. |
| first_indexed | 2024-03-10T01:00:18Z |
| format | Article |
| id | doaj.art-162ca2fa2c8241dfb80be56215d77991 |
| institution | Directory Open Access Journal |
| issn | 2073-4395 |
| language | English |
| last_indexed | 2024-03-10T01:00:18Z |
| publishDate | 2022-08-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Agronomy |
| spelling | doaj.art-162ca2fa2c8241dfb80be56215d779912023-11-23T14:36:04ZengMDPI AGAgronomy2073-43952022-08-01129202310.3390/agronomy12092023DS-DETR: A Model for Tomato Leaf Disease Segmentation and Damage EvaluationJianshuang Wu0Changji Wen1Hongrui Chen2Zhenyu Ma3Tian Zhang4Hengqiang Su5Ce Yang6College of Information and Technology, Jilin Agricultural University, Changchun 130118, ChinaCollege of Information and Technology, Jilin Agricultural University, Changchun 130118, ChinaCollege of Information and Technology, Jilin Agricultural University, Changchun 130118, ChinaCollege of Information and Technology, Jilin Agricultural University, Changchun 130118, ChinaCollege of Information and Technology, Jilin Agricultural University, Changchun 130118, ChinaCollege of Information and Technology, Jilin Agricultural University, Changchun 130118, ChinaCollege of Food, Agricultural and Natural Resource Sciences, University of Minnesota, St. Paul, MN 55108, USAEarly blight and late blight are important factors restricting tomato yield. However, it is still a challenge to accurately and objectively detect and segment crop diseases in order to evaluate disease damage. In this paper, the Disease Segmentation Detection Transformer (DS-DETR) is proposed to segment leaf disease spots efficiently based on several improvements to DETR. Additionally, a damage assessment is carried out by the area ratio of the segmented leaves to the disease spots. First, an unsupervised pre-training method was introduced into DETR with the Plant Disease Classification Dataset (PDCD) to solve the problem of the long training epochs and slow convergence speed of DETR. This method can train the Transformer structures in advance to obtain leaf disease features. Loading the pre-training model weight in DS-DETR can speed up the convergence speed of the model. Then, Spatially Modulated Co-Attention (SMCA) was used to assign Gaussian-like spatial weights to the query box of DS-DETR. The different positions in the image are trained using the query boxes with different weights to improve the accuracy of the model. Finally, an improved relative position code was added to the Transformer structure of DS-DETR. Relative position coding promotes the capture of the sequence order of input tokens by the Transformer. The spatial location feature is strengthened by establishing the location relationship between different instances. Based on these improvements, the DS-DETR model was tested on the Tomato leaf Disease Segmentation Dataset (TDSD) constructed by us. The experimental results show that the DS-DETR proposed by us achieved 0.6823 for AP<sub>mask</sub>, which improved by 12.87%, 8.25%, 3.67%, 1.95%, 10.27%, and 9.52% compared with the state-of-the-art: Mask RCNN, BlendMask, CondInst, SOLOv2, ISTR, and DETR, respectively. In addition, the disease grading accuracy reached 0.9640 according to the segmentation results given by our proposed model.https://www.mdpi.com/2073-4395/12/9/2023tomato diseasedisease damage evaluationdetection transformerdeep learninginstance segmentation |
| spellingShingle | Jianshuang Wu Changji Wen Hongrui Chen Zhenyu Ma Tian Zhang Hengqiang Su Ce Yang DS-DETR: A Model for Tomato Leaf Disease Segmentation and Damage Evaluation Agronomy tomato disease disease damage evaluation detection transformer deep learning instance segmentation |
| title | DS-DETR: A Model for Tomato Leaf Disease Segmentation and Damage Evaluation |
| title_full | DS-DETR: A Model for Tomato Leaf Disease Segmentation and Damage Evaluation |
| title_fullStr | DS-DETR: A Model for Tomato Leaf Disease Segmentation and Damage Evaluation |
| title_full_unstemmed | DS-DETR: A Model for Tomato Leaf Disease Segmentation and Damage Evaluation |
| title_short | DS-DETR: A Model for Tomato Leaf Disease Segmentation and Damage Evaluation |
| title_sort | ds detr a model for tomato leaf disease segmentation and damage evaluation |
| topic | tomato disease disease damage evaluation detection transformer deep learning instance segmentation |
| url | https://www.mdpi.com/2073-4395/12/9/2023 |
| work_keys_str_mv | AT jianshuangwu dsdetramodelfortomatoleafdiseasesegmentationanddamageevaluation AT changjiwen dsdetramodelfortomatoleafdiseasesegmentationanddamageevaluation AT hongruichen dsdetramodelfortomatoleafdiseasesegmentationanddamageevaluation AT zhenyuma dsdetramodelfortomatoleafdiseasesegmentationanddamageevaluation AT tianzhang dsdetramodelfortomatoleafdiseasesegmentationanddamageevaluation AT hengqiangsu dsdetramodelfortomatoleafdiseasesegmentationanddamageevaluation AT ceyang dsdetramodelfortomatoleafdiseasesegmentationanddamageevaluation |