This document discusses several papers on tomato disease detection and grading using computer vision techniques:
1. The first paper uses histogram thresholding and an RBF-SVM classifier to identify stalk scars and calyx on tomatoes with 95.15% accuracy. It achieves overall grading accuracy but this decreases with more grading levels.
2. The second paper uses a whale optimization based artificial neural network to identify multiple tomato diseases with better accuracy than other models. It evaluates performance using metrics like sensitivity, specificity, and F1 score.
3. The third paper discusses an 8-layer CNN model that achieves 98.4% accuracy on the plant village dataset for tomato disease classification, outperforming other models. It
This document discusses several papers on tomato disease detection and grading using computer vision techniques:
1. The first paper uses histogram thresholding and an RBF-SVM classifier to identify stalk scars and calyx on tomatoes with 95.15% accuracy. It achieves overall grading accuracy but this decreases with more grading levels.
2. The second paper uses a whale optimization based artificial neural network to identify multiple tomato diseases with better accuracy than other models. It evaluates performance using metrics like sensitivity, specificity, and F1 score.
3. The third paper discusses an 8-layer CNN model that achieves 98.4% accuracy on the plant village dataset for tomato disease classification, outperforming other models. It
This document discusses several papers on tomato disease detection and grading using computer vision techniques:
1. The first paper uses histogram thresholding and an RBF-SVM classifier to identify stalk scars and calyx on tomatoes with 95.15% accuracy. It achieves overall grading accuracy but this decreases with more grading levels.
2. The second paper uses a whale optimization based artificial neural network to identify multiple tomato diseases with better accuracy than other models. It evaluates performance using metrics like sensitivity, specificity, and F1 score.
3. The third paper discusses an 8-layer CNN model that achieves 98.4% accuracy on the plant village dataset for tomato disease classification, outperforming other models. It
al [1] In this paper they discussed RGB images based on the
tomato grading vision-based system. They proposed the idea accomplishing stalk scar identification and calyx at a good accuracy level, i.e, 0.9515 for both the healthy and infected tomatoes by applying histogram thresholding that depends on the g-r value of these area of interest. However defected area is identified by using a RBF- SVM classifier by applying the LAB colour space pixel value. Therefore, the models attain an overall accuracy consequent to validation. Generally, in this paper they are using four grading levels that were refined based on texture and colour. To better understand, the healthy and infected category, the RBF-SVM exceeds all the research models with the highest accuracy of 0.9709. After that when the number of grading levels increased then the grading accuracy is decreased. However, the closer observation is that they are using an inline tomato sorting tool to secure that the quality is maintained. S.Dhakshina, S.Esakkirajan, C.veemalraj, et.al[2] According to this author, so many methods were applied for identification of diseases in plants. The overall observation is that successful identification with better accuracy was not attain in any of the developed models. Therefore, for better prediction of several types of detection of diseases in tomatoes they are using Whale Optimization Based Artificial Neural Network. However, to calculate the performance of the nominated system some of the analytical parameters are sensitivity, specificity, F1 score, error rate, accuracy. Except from improvements in performance of the classification and prediction of diseases in tomato which is compared to the other convolutional method like, Artificial Neural Network, K- Nearest Neighbour, Probabilistic Neural Network, Back Propagation Artificial Neural Network, a differentiation evaluation is performed. Agarwal, Gupta, Biswas et.al[3] had discussed a simplified CNN model which suggested it included 8 hidden layers. By applying the openly available plant village dataset. After that, it proposed a light-weight model, performed better as compared to conventional machine learning approaches as well as related models accomplish an accuracy of 98.4%. Generally, the village dataset consists of 39 classes of several crops like potato, corn, apple, grapes etc. It is interesting to see that there are 10 classes of tomato diseases. Whereas convolutional machine learning models give better results of 94.9% with K-NN, better accuracy of 93.5% is attain to VGG 16 in related models. Moreover, to increment the performance of suggested CNN for changing the image brightness, image processing is used with a random value of image random width after using image augmentation. The suggested model performs greatly well on the dataset other than the plant village dataset with best accuracy of 98.7%. Mkonyi, Rubanga, Ruchard, Zekeya et.al[4] In this paper the author gives an idea to build up phytosanitary capacity and to assist tomato plant pest tuta absolute destruction at the very growing stage of the tomato. However, applying deep learning technique to identify the disease. Whereas for training classifier on tomato image dataset that is caught from the field which includes healthy and sick leaves by using Convolutional neural network models like (VGG 16, VGG19 and ResNet 50). They calculate performance for every classifier by considering the accuracy for categorize the tomato canopy in the correct group. We seek to identify that the experimental results that are VGG 16 acquires the highest accuracy value of 91.9%. Chen, Zhou, Yi, Zhang et.al [5] author suggested that they proposed a new methodology for identification of disease. However, the image is denoised and then improved by Binary Wavelet Transform with Retinex (BWTR). Whereas the noise point and edge point are removed and the most important texture information is continued. Generally, then the tomatoes were splitted from the background by applying KSW optimized by using Artificial Bee Colony Algorithm. Subsequently, the both-channel Residual Attention Network model (B-ARNet) was used to detect the image. However, the overall exposure accuracy is about 89.9%. The overall contribution of this paper is that tomato leaf disease detection method depends on the mixture of B-ARNet, BWTR, ABCK it is efficient.