A novel two-stage method of plant seedlings classification based on deep learning

2021 ◽  
pp. 1-11
Author(s):  
Tianhong Dai ◽  
Shijie Cong ◽  
Jianping Huang ◽  
Yanwen Zhang ◽  
Xinwang Huang ◽  
...  

In agricultural production, weed removal is an important part of crop cultivation, but inevitably, other plants compete with crops for nutrients. Only by identifying and removing weeds can the quality of the harvest be guaranteed. Therefore, the distinction between weeds and crops is particularly important. Recently, deep learning technology has also been applied to the field of botany, and achieved good results. Convolutional neural networks are widely used in deep learning because of their excellent classification effects. The purpose of this article is to find a new method of plant seedling classification. This method includes two stages: image segmentation and image classification. The first stage is to use the improved U-Net to segment the dataset, and the second stage is to use six classification networks to classify the seedlings of the segmented dataset. The dataset used for the experiment contained 12 different types of plants, namely, 3 crops and 9 weeds. The model was evaluated by the multi-class statistical analysis of accuracy, recall, precision, and F1-score. The results show that the two-stage classification method combining the improved U-Net segmentation network and the classification network was more conducive to the classification of plant seedlings, and the classification accuracy reaches 97.7%.

Author(s):  
Mohammad Rizk Assaf ◽  
Abdel-Nasser Assimi

In this article, the authors investigate the enhanced two stage MMSE (TS-MMSE) equalizer in bit-interleaved coded FBMC/OQAM system which gives a tradeoff between complexity and performance, since error correcting codes limits error propagation, so this allows the equalizer to remove not only ICI but also ISI in the second stage. The proposed equalizer has shown less design complexity compared to the other MMSE equalizers. The obtained results show that the probability of error is improved where SNR gain reaches 2 dB measured at BER compared with ICI cancellation for different types of modulation schemes and ITU Vehicular B channel model. Some simulation results are provided to illustrate the effectiveness of the proposed equalizer.


Electronics ◽  
2021 ◽  
Vol 10 (4) ◽  
pp. 495
Author(s):  
Imayanmosha Wahlang ◽  
Arnab Kumar Maji ◽  
Goutam Saha ◽  
Prasun Chakrabarti ◽  
Michal Jasinski ◽  
...  

This article experiments with deep learning methodologies in echocardiogram (echo), a promising and vigorously researched technique in the preponderance field. This paper involves two different kinds of classification in the echo. Firstly, classification into normal (absence of abnormalities) or abnormal (presence of abnormalities) has been done, using 2D echo images, 3D Doppler images, and videographic images. Secondly, based on different types of regurgitation, namely, Mitral Regurgitation (MR), Aortic Regurgitation (AR), Tricuspid Regurgitation (TR), and a combination of the three types of regurgitation are classified using videographic echo images. Two deep-learning methodologies are used for these purposes, a Recurrent Neural Network (RNN) based methodology (Long Short Term Memory (LSTM)) and an Autoencoder based methodology (Variational AutoEncoder (VAE)). The use of videographic images distinguished this work from the existing work using SVM (Support Vector Machine) and also application of deep-learning methodologies is the first of many in this particular field. It was found that deep-learning methodologies perform better than SVM methodology in normal or abnormal classification. Overall, VAE performs better in 2D and 3D Doppler images (static images) while LSTM performs better in the case of videographic images.


2021 ◽  
Author(s):  
Mizuho Mori ◽  
Yoshiko Ariji ◽  
Motoki Fukuda ◽  
Tomoya Kitano ◽  
Takuma Funakoshi ◽  
...  

Abstract Objectives The aim of the present study was to create and test an automatic system for assessing the technical quality of positioning in periapical radiography of the maxillary canines using deep learning classification and segmentation techniques. Methods We created and tested two deep learning systems using 500 periapical radiographs (250 each of good- and bad-quality images). We assigned 350, 70, and 80 images as the training, validation, and test datasets, respectively. The learning model of system 1 was created with only the classification process, whereas system 2 consisted of both the segmentation and classification models. In each model, 500 epochs of training were performed using AlexNet and U-net for classification and segmentation, respectively. The segmentation results were evaluated by the intersection over union method, with values of 0.6 or more considered as success. The classification results were compared between the two systems. Results The segmentation performance of system 2 was recall, precision, and F measure of 0.937, 0.961, and 0.949, respectively. System 2 showed better classification performance values than those obtained by system 1. The area under the receiver operating characteristic curve values differed significantly between system 1 (0.649) and system 2 (0.927). Conclusions The deep learning systems we created appeared to have potential benefits in evaluation of the technical positioning quality of periapical radiographs through the use of segmentation and classification functions.


2021 ◽  
Vol 13 (7) ◽  
pp. 1236
Author(s):  
Yuanjun Shu ◽  
Wei Li ◽  
Menglong Yang ◽  
Peng Cheng ◽  
Songchen Han

Convolutional neural networks (CNNs) have been widely used in change detection of synthetic aperture radar (SAR) images and have been proven to have better precision than traditional methods. A two-stage patch-based deep learning method with a label updating strategy is proposed in this paper. The initial label and mask are generated at the pre-classification stage. Then a two-stage updating strategy is applied to gradually recover changed areas. At the first stage, diversity of training data is gradually restored. The output of the designed CNN network is further processed to generate a new label and a new mask for the following learning iteration. As the diversity of data is ensured after the first stage, pixels within uncertain areas can be easily classified at the second stage. Experiment results on several representative datasets show the effectiveness of our proposed method compared with several existing competitive methods.


The increased usage of the Internet and social networks allowed and enabled people to express their views, which have generated an increasing attention lately. Sentiment Analysis (SA) techniques are used to determine the polarity of information, either positive or negative, toward a given topic, including opinions. In this research, we have introduced a machine learning approach based on Support Vector Machine (SVM), Naïve Bayes (NB) and Random Forest (RF) classifiers, to find and classify extreme opinions in Arabic reviews. To achieve this, a dataset of 1500 Arabic reviews was collected from Google Play Store. In addition, a two-stage Classification process was applied to classify the reviews. In the first stage, we built a binary classifier to sort out positive from negative reviews. In the second stage, however we applied a binary classification mechanism based on a set of proposed rules that distinguishes extreme positive from positive reviews, and extreme negative from negative reviews. Four major experiments were conducted with a total of 10 different sub experiments to fulfill the two-stage process using different X-validation schemas and Term Frequency-Inverse Document Frequency feature selection method. Obtained results have indicated that SVM was the best during the first stage classification with 30% testing data, and NB was the best with 20% testing data. The results of the second stage classification indicated that SVM has scored better results in identifying extreme positive reviews when dealing with the positive dataset with an overall accuracy of 68.7% and NB showed better accuracy results in identifying extreme negative reviews when dealing with the negative dataset, with an overall accuracy of 72.8%.


IEEE Access ◽  
2019 ◽  
Vol 7 ◽  
pp. 41770-41781 ◽  
Author(s):  
Catherine Sandoval ◽  
Elena Pirogova ◽  
Margaret Lech

2018 ◽  
Vol 5 (2) ◽  
pp. 102
Author(s):  
Dede Zaenal Arif

The purpose of this research is to produce fish crackers from different types of fish, namely catfish and patin fish as well as different types of starch and know the characteristics of a good fish crackers. The benefit that can be expected from this research was to utilize catfish and patin fish abundant potency and add economic value. The method of this research was divided into two stages, namely the first stage is the stage which determines the range of the comparison with the fish flour, determine the type of fish and determine the type of flour used by using the hedonik method of organoleptic parameters. On the second stage has a purpose and that is to analyze chemical and physical fish crackers by comparison. The data were analyzed using the method of experiment results simple linear variable (x) increase in comparison of fish and flour (part). The free variable (y) consists of the response of the color, flavor, aroma, texture, volume and the development levels of crispness. The type of fish and the type of starch correlated against all response organoleptic, except the catfish and tapioca flour was not correlated against sense, catfish and cornmeal were not correlated against the texture of the fish, and catfish and sago flour not correlated against scent. The highest correlation is indicated by the sample composition of the cornmeal and catfish fish total value index by 17 of the total value of the correlation coefficient in classification. The sample was selected based on organoleptic level consumer favorite is with the composition of samples catfish and tapioca flour with a 1:1 comparison (111). Based on the results of the chemical analysis of protein obtained 24,38%, fat content of 1.6%, levels of starch of 44.69% and water content of 5.5%. Physical analysis of the parameter and the mobilising of the volume development of IE of 146.43% and the level of crispness that is of 0.56 mm/s/50gram.  


2019 ◽  
Vol 54 (S1) ◽  
pp. 86-87
Author(s):  
X.P. Burgos‐Artizuu ◽  
E. Eixarch ◽  
D. Coronado‐Gutierrez ◽  
B. Valenzuela ◽  
E. Bonet‐Carne ◽  
...  

2020 ◽  
Vol 10 (19) ◽  
pp. 6866
Author(s):  
Arnauld Nzegha Fountsop ◽  
Jean Louis Ebongue Kedieng Fendji ◽  
Marcellin Atemkeng

Deep learning has been successfully showing promising results in plant disease detection, fruit counting, yield estimation, and gaining an increasing interest in agriculture. Deep learning models are generally based on several millions of parameters that generate exceptionally large weight matrices. The latter requires large memory and computational power for training, testing, and deploying. Unfortunately, these requirements make it difficult to deploy on low-cost devices with limited resources that are present at the fieldwork. In addition, the lack or the bad quality of connectivity in farms does not allow remote computation. An approach that has been used to save memory and speed up the processing is to compress the models. In this work, we tackle the challenges related to the resource limitation by compressing some state-of-the-art models very often used in image classification. For this we apply model pruning and quantization to LeNet5, VGG16, and AlexNet. Original and compressed models were applied to the benchmark of plant seedling classification (V2 Plant Seedlings Dataset) and Flavia database. Results reveal that it is possible to compress the size of these models by a factor of 38 and to reduce the FLOPs of VGG16 by a factor of 99 without considerable loss of accuracy.


Sign in / Sign up

Export Citation Format

Share Document