Impact of dataset size and variety on the effectiveness of deep learning and transfer learning for plant disease classification

Barbedo, Jayme Garcia Arnal

doi:10.1016/j.compag.2018.08.013

Cited by 482 publications

(259 citation statements)

References 14 publications

Supporting

Mentioning

245

Contrasting

Unclassified

Order By: Relevance

“…Data diversity is one of the key factors to ensure the generalization ability of the model [37]. Although the study used nearly 10,000 images, the amount of data was still small for 10 soybean varieties.…”

Section: Comparison Analysismentioning

confidence: 99%

“…In addition, if the seeds are not severely wet (seed color does not change and seeds do not swell), the relative humidity has no influence on the method of hyperspectral imaging combined deep learning. At present, deep networks have been successfully applied to plant disease identification [36][37][38], drought monitoring [39], land type classification [40], weed detection [41], and other areas of agriculture. To date, there are few reports on the identification of soybean seed varieties by deep learning, and whether it has advantages that is also unknown.…”

mentioning

confidence: 99%

See 1 more Smart Citation

A Rapid and Highly Efficient Method for the Identification of Soybean Seed Varieties: Hyperspectral Images Combined with Transfer Learning

et al. 2019

View full text Add to dashboard Cite

Convolutional neural network (CNN) can be used to quickly identify crop seed varieties. 1200 seeds of ten soybean varieties were selected, hyperspectral images of both the front and the back of the seeds were collected, and the reflectance of soybean was derived from the hyperspectral images. A total of 9600 images were obtained after data augmentation, and the images were divided into a training set, validation set, and test set with a 3:1:1 ratio. Pretrained models (AlexNet, ResNet18, Xception, InceptionV3, DenseNet201, and NASNetLarge) after fine-tuning were used for transfer training. The optimal CNN model for soybean seed variety identification was selected. Furthermore, the traditional machine learning models for soybean seed variety identification were established by using reflectance as input. The results show that the six models all achieved 91% accuracy in the validation set and achieved accuracy values of 90.6%, 94.5%, 95.4%, 95.6%, 96.8%, and 97.2%, respectively, in the test set. This method is better than the identification of soybean seed varieties based on hyperspectral reflectance. The experimental results support a novel method for identifying soybean seeds rapidly and accurately, and this method also provides a good reference for the identification of other crop seeds.

show abstract

Section: Comparison Analysismentioning

confidence: 99%

mentioning

confidence: 99%

A Rapid and Highly Efficient Method for the Identification of Soybean Seed Varieties: Hyperspectral Images Combined with Transfer Learning

et al. 2019

View full text Add to dashboard Cite

show abstract

“…Training beyond 10 epochs did not improve the accuracy hence it was stopped as it may result in overfitting. When the model is excessively trained, it memorizes the patterns of the training dataset leading to a poor generalization 15,30 . The architecture trained with RGB images resulted in a maximum mean classification accuracy of 95.1% and when the proportion of samples for each class was taken into consideration, the accuracy was 94.7% as shown in Fig.…”

Section: Resultsmentioning

confidence: 99%

“…Hence there is a lack of availability of large disease dataset which is a potential area for improvement. Also, it demands a costly system equipped with a Graphics Processing Unit (GPU) and large Random Access Memory (RAM) for training the models 12,13,15 . Some of the studies used an approach called transfer learning approach where the pre-trained models have been used for disease classification 1,12,14-17 .…”

mentioning

confidence: 99%

Disease Classification in Eggplant Using Pre-trained VGG16 and MSVM

Rangarajan

Raja

2020

Sci Rep

141

View full text Add to dashboard Cite

Currently, the application of deep learning in crop disease classification is one of the active areas of research for which an image dataset is required. Eggplant (Solanum melongena) is one of the important crops, but it is susceptible to serious diseases which hinder its production. Surprisingly, so far no dataset is available for the diseases in this crop. The unavailability of the dataset for these diseases motivated the authors to create a standard dataset in laboratory and field conditions for five major diseases. Pre-trained Visual Geometry Group 16 (VGG16) architecture has been used and the images have been converted to other color spaces namely Hue Saturation Value (HSV), YCbCr and grayscale for evaluation. Results show that the dataset created with RGB and YCbCr images in field condition was promising with a classification accuracy of 99.4%. The dataset also has been evaluated with other popular architectures and compared. In addition, VGG16 has been used as feature extractor from 8 th convolution layer and these features have been used for classifying diseases employing Multi-Class Support Vector Machine (MSVM). The analysis depicted an equivalent or in some cases produced better accuracy. Possible reasons for variation in interclass accuracy and future direction have been discussed.The revolution of the modern technologies in the recent era has facilitated its application in agriculture to improve production. One of the applications is the diagnosis of plant diseases using a digital image from a camera which in turn will assist the farmers to control its prevalence in the fields. The availability of cheap cameras and the explosive growth on the internet have made the diagnosis relatively less complex with the availability of tools and information about the disease online 1 . But still, human diagnosis is prone to errors 2 . The scope for the automatic disease classification has improved due to the accomplishment in machine learning technologies. Traditionally shallow machine learning algorithms such as neural networks, Support Vector Machine (SVM), or other algorithms were used which is a time-consuming process as it demands feature extraction from the images manually and fed as input to the algorithm for classification. But, the deep learning approaches consist of many layers of processing elements that process images and estimate features automatically for classification. There are four major types of deep learning algorithms namely Convolutional Neural Networks (CNN), autoencoder, restricted Boltzmann machines and sparse encoding, according to a study by Guo et al. 3 Of these, CNN based architectures are most widely used for image classification problems 3 . Recent trends in the use of CNN for disease classification are on the rise and many studies have reported promising results 1-17 .Training of the CNN based deep learning models from scratch is a time consuming (difficult) process and requires a large database. It is also challenging to categorize each image to a crop disease even with an expert...

show abstract

“…However, computer vision based machine learning algorithms-those which are able to identify objects in images-require a large amount of training data to be able to recognise any specific object. The accuracy of these algorithms deteriorates rapidly if insufficient training data are available [26][27][28], and, in some cases, simpler statistical methods have been shown to perform better than deep learning algorithms for object recognition [29].…”

Section: Introductionmentioning

confidence: 99%

Requirements and Limitations of Thermal Drones for Effective Search and Rescue in Marine and Coastal Areas

Burke

McWhirter

Veitch-Michaelis

et al. 2019

Drones

View full text Add to dashboard Cite

Search and rescue (SAR) is a vital line of defense against unnecessary loss of life. However,in a potentially hazardous environment, it is important to balance the risks associated with SARaction. Drones have the potential to help with the efficiency, success rate and safety of SAR operationsas they can cover large or hard to access areas quickly. The addition of thermal cameras to the dronesprovides the potential for automated and reliable detection of people in need of rescue. We performeda pilot study with a thermal-equipped drone for SAR applications in Morecambe Bay. In a varietyof realistic SAR scenarios, we found that we could detect humans who would be in need of rescue,both by the naked eye and by a simple automated method. We explore the current advantages andlimitations of thermal drone systems, and outline the future path to a useful system for deploymentin real-life SAR.

show abstract

Impact of dataset size and variety on the effectiveness of deep learning and transfer learning for plant disease classification

Cited by 482 publications

References 14 publications

A Rapid and Highly Efficient Method for the Identification of Soybean Seed Varieties: Hyperspectral Images Combined with Transfer Learning

A Rapid and Highly Efficient Method for the Identification of Soybean Seed Varieties: Hyperspectral Images Combined with Transfer Learning

Disease Classification in Eggplant Using Pre-trained VGG16 and MSVM

Requirements and Limitations of Thermal Drones for Effective Search and Rescue in Marine and Coastal Areas

Contact Info

Product

Resources

About