CN7023 Assignment Sample – Artificial Intelligence & Machine Vision 2022
CLASSIFICATION OF FRUIT IMAGES USING THE CONVENTIONAL NEURAL NETWORK
Abstract : The main aim of this report is considered with the selection of neural network-based algorithm according to the service management based on the selected data set. Here considered with the fruits data sets that is contain the information of fruits and vegetables based on that the result of obtained.
For gathering the output here implementation is done using the MATLAB and create the simulations based on the testing dataset and training dataset according to the service and process analysis of the measurement.
This is can be implemented using the various analysis according to the specification of measurement based on the analysis that is considered with the effective mapping and measuring the correct value according to the service enhancement process.
Here the accuracy of the selected dataset is find based on the implementation on the MATLAB using the conventional Neural Network process.
1.Introduction
The main objective of this course work is focus on the analysis of selected dataset and then it is implemented in the MATLAB according to the selection of Conventional Neural Network. This is focus on the analysis of traditional machine learning algorithms according to the different kind of measurement based on the selected value. It is considered with the set limitation along with the number of samples that is used in the research.
The conventional Neural Network is developed for managing the classification of images and provide the accurate result of classification according to the measurement for the analysis and process of measurement. It can be classified class 1, class2 etc.
The classification is done based on the statistical classifier and probabilistic classifier according to the functionality of various image segregation process. This is considered with the artificial neural network for the classification of the result based on the measurement of different kind of applications according to the specifications.
The research question is providing the answer for the different perspective of result analysis.
- How to segregate the image based on the attributes and based on the position of image?
- Is any different classifiers is available for the image and text?
- Conventional Neural Network is providing the efficient result according to the various analysis of dataset.
- What are the fields need to measure while done the classification according to the specification of the process?
- It may change in future if better algorithm is used for the classification
Like that it is focus on the different dimension of the analysis according to the fruit dataset based on the measurement and provide the efficient result according to the algorithm implementation in the MATLAB for the efficient mapping of attributes of the images.
-
Methodology
The main set of the machine learning needs to be in the twist of the recent activities has an advantage of the new rise in the artificial intelligence have the need of the network has an need of the models are to be in the set of the common set of the machine learning has some of the tasks to be done in the set of the computational value which is need to be done in the value to arrange the need of the impressive form has .
ANN which is used for the solve the different need of the image has an effect with the image driven pattern has an task to get recognized in the task as set of the simple architecture to be done with the sampled method has an need of the introduction of the convolutional need of the image has an recognition which is used for the fundamentals as well as the machine learning.
Has a need of an application of the learning set of the conventional set of operation to be done in the side of the parameter have an outside of the conventional layer such as the kernel which is connected with the terminology, CNN has an model for an deep learning has an need for the processing information such as the data.
The traditional set of the Convolutional set of the neurons which is used for the self-optimize has an learning set of the neurons which is used for the development of an input has an need of the performance of the operation has an need of the need of the product has an function of non-linear has the basic set of the countless has an need of the CNN ,.
The following terms said by the author described the employed having the parameters to be in the set of variables as well as the automation has to done with the different set of the training process as well as the hyperparameters will said will be the need of the training set of the process state.
Has an input has an need of the raw kind of image which is need for the output has an function to get in the set of the network to work have the layer for some of the function to score the basic layer which have associated with the regular tips has to be done with the traditional CNN .
The layer of convolutional layer have the side of the fully connected layer has an need of the function has an set of the model to be done with the performance to be done with the set of the particular kernel as well as the weight to be done.
With the set of the function to get loss has an need of the forward propagation has to done with the set of the training dataset has an need of the set of the kernel as well as the weight has an need of the loss of value have the back propagation with the decent set of the optimization to follow in the set of the algorithm to refined in the rectified linear unit [1].
The main set of the notable set has a difference in the set of the CNNs as well as the traditional kind have to done with the set of the field to be include in the pattern has to be recognized in the process of image which is used to allow the encode of the image which is used for the more suit has an need of the focused set of the tasks which is need of the parameters.
One of the largest set of the traditional set of the structure has an need of the largest set of the traditional form have to compute the data has an need of the normalized in the set of the management has an substance need to be done with the side of the colored structure has an need of the weight has an single mode.
Account has an need to clarify the set has an effect has an classify the color normalized for the digit having the layer used for the function to don with the factor of three dimension has the input of some height as well as the width and the depth of some function to be included in the set of the CNN [2].
The neurons has to make the better the layer has and small set of the region has an understand of the layer having the input of some dimensional which would have the scroll of the system to be done with the side of the depth of some dimension to be included in the set of the pooling layer has an layer used for the CNN .
Has an need of the grid which has the image which is connected with the organizational as well as the need of the visual set which is connected with the low – high level having the pattern to be done with the case of the mathematical function is to connect with the three layers as well as it need to be done with the side of the some type of layers such as the combining as well as the layers to be connected.
The first two are forms to perform the function to be done with the side of the feature extraction as well as the final output will get as a connection called as the classification of the connected layer , the conventional layer said to be the form of the stack have in the set of the composed form of the mathematical operational which is said to be in the need of the set of the linear set of the linear operation has to perform the image processing in the set of the pixel side of the values which is stored in the value of the.
Two dimensional grid having the function to be done in the side of the array having the number of arrays to be included in the grid having the functionalities has an need of the grid having the parameters to be included in the set of the image processing has an specific set of the output want to reach in the set of the layer have to be done with the output which has an layer to be done with the set of the feature has an progress as well as the hierarchical has to done with the more complex set.
The function which is made by the sample of the example of CNN has an applied set of the broken down as an found in the set of the forms has an CNN which is set to be found on the input layer which will hold the direct level has an need of the convolutional layer has an output of some neurons to get the basic factor have an calculation to do it in the side of the volume of an input [3] .,
The linear unit of an rectified set of the activation to be done with the output to be done in the case of the previous layer. So most of the special as well as the process of the optimization to be done with the side of the kernel in the set of the kernel which is want to perform have the minimization to become more different in the need of the output as well as it connected with the set of the ground level has to done with the algorithm of the optimization which is called as the gradient as well as the backpropagation.
Has an some overview has an some convolutional neural network has an advantage of the architecture as well as the training aspects of the training process as well as the CNN has an several need of the building block of system set in the conventional neural network [4].
-
Simulations
The simulation result is focus on the implementation of the Fruits dataset.
3.1 Dataset
The dataset used for this analysis fruit based and it is considered with the following attributes. It is gathered from the given link https://www.kaggle.com/moltean/fruits.
The Fruit dataset is contain the total number of image is 90483. In this the training dataset is considered with the size of 67692 images and the testing dataset is considered with the size of 22688 images. The total number of classes is considered with the 131 classes.
The file format may be in the JPEG. The representation of r2 is represented with the fruit that was rotated in the format then it is focus on the various analysis of the process according to the measurement of functionality according to the service. The different variety of same fruits are stored based on different classes. The pictures are stored in the multiple folders according to the measurement based on the utilization of each access.
Figure 1: Image Classification
The maximum value of images are stored in the prescribed format according to the verification and measurement. Like that dataset is considered for the next kind of analysis of image classification based on the measurement of each attributes of images according to the service enhancement according to the specification.
3.2 Network Architecture
The network architecture is created based on the measurement of Conventional Neural network process and this can be considered with the position of the image and properties of images according to the service and utilization based on the analysis.
Figure 2: Structure of Convolutional Neural Network
According to the Conventional Neural Network is considered with the effective mapping and process analysis according to the measurement of images. The CNN is considered with the type of neural network that is used in the image classification based on the pixel of data according to the measurement of pixel value it is focus on the various analysis based on the different kind of applications according to the service process. this is used with the effective analysis according to the specification.
It is focus on the image classifiers based on the position of image that is analysis using the CNN value according to the services. This is focus on the different kind of measurement of images along with the specification of pixels according to the analysis for the effective mapping.
Figure 3: Fruits Types
The result is obtained based on the selected given images according to the analysis of various process then it is focus on the images position in the tool then it is mapping into the collection dataset that is mapped into the specification of various analysis based on the measurement of different kind of analysis for the betterment of image classification.
This is considered with the effective mapping according to the analysis result provide the effective mapping system according to the measurement and analysis.
-
Results Obtained
The CNN methodology is considered with the Fruit dataset that is considered with the following classification results. This is focus on the training dataset results and testing dataset result according to the various measurement. It is considered with the effective measurement for the analysis of the of the process. It is considered with the effective measurement of process according to the use of analysis because of utilization.
Figure 4: Overall Training Accuracy
This can be process with the effective mapping according to the service enhancement. The above graph is showing the result of process according to the services. The result shows the information of overall training process accuracy 92.7% according to the specification of the measurement based on the analysis of each process.
it is find based on each epoch in the scenario according to the finding result. t is also focus on the different kind of image and process that is used by the CNN according to the various analysis.
Figure 5: Overall Testing Accuracy
In this scenario it is focus on the measurement of process according to the result obtained based on the accuracy implementation of effective result. The result shows the information of overall testing process accuracy 92.7% according to the specification of the measurement based on the analysis of each process. it is find based on each epoch in the scenario according to the finding result.
Figure 6: Confusion Matrix
The confusion matrix is generated according to each epoch based on the three different classifiers according to the measurement of various analysis. It is showing the target class and output class for the betterment of usage according to the scene of various applications. This can be enhancing and accrue the percentage of analysis according to the process and measurement.
-
Critical analysis of result
The process of analysis the result according to the overall testing accuracy of the dataset is gained 96.2% and overall training accuracy of the dataset is gained with the 96.2%. it is focus on the different kind of image and process that is used by the CNN according to the various analysis. In this scenario it is focus on the measurement of process according to the result obtained based on the accuracy implementation of effective result.
The Fruit dataset is containing the total number of images is 90483. In this the training dataset is considered with the size of 67692 images and the testing dataset is considered with the size of 22688 images based on that it is focus on the result obtained.
The result is obtained based on the selected given images according to the analysis of various process then it is focus on the images position in the tool then it is mapping into the collection dataset that is mapped into the specification of various analysis.
-
Conclusion
The analysis of selected dataset and then it is implemented in the MATLAB according to the selection of Conventional Neural Network. The Fruit dataset is containing the total number of images is 90483. In this the training dataset is considered with the size of 67692 images and the testing dataset is considered with the size of 22688 images.
The total number of classes is considered with the 131 classes. The process of analysis the result according to the overall testing accuracy of the dataset is gained 96.2% and overall training accuracy of the dataset is gained with the 96.2%. it is focus on the different kind of image and process that is used by the CNN according to the various analysis.
References
- Benz, P. Hofmann, G. Willhauck, I. Lingenfelder, and M. Heynen, “Multi-resolution, object-oriented fuzzy analysis of remote sensing data for GIS-ready information,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 58, pp. 239–258, 2004.
- Blaschke, “Object-based image analysis for remote sensing,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 65, no. 1, pp. 2–16, 2010.
- Li, L. Ni, X. Jia, L. Gao, B. Zhang, and M. Peng, “Multi-scale superpixel spectral-spatial classification of hyperspectral images,” International journal of remote sensing, vol. 37, no. 19–20, pp. 4905–4922, 2016.
W.S. McCulloch and W. Pitts, “A logical calculus of the ideas immanent in nervous activity,” Bulletin of mathematical biology, vol. 5, no. 4, pp. 115–133, 1943.
- Martine and T. Jean-Bernard, “Neural approach for TV image compression using a Hopfield type network,” in Advances in Neural Information Processing Systems 1, D.S. Touretzky, Ed., pp. 264–271. 1989.
- Blaschke, “Object-based image analysis for remote sensing,” ISPRS Journal of Photogrammetry and Remote Sensing, vol. 65, no. 1, pp. 2–16, 2010.
- Li, L. Ni, X. Jia, L. Gao, B. Zhang, and M. Peng, “Multi-scale superpixel spectral-spatial classification of hyperspectral images,” International journal of remote sensing, vol. 37, no. 19–20, pp. 4905–4922, 2016.
- Zhang, S. Zhang, G. Sun, F. Li, and Z. Wang, “Mapping of coastal cities using optimized spectral–spatial features based multi-scale superpixel classification,” Remote Sensing, vol. 11, no. 9, 2019
Hamid, S. S. A. A. N. M. N. K. A. &. A. G. A. A.,. “Dyslexia adaptive learning model: student engagement prediction using machine learning approach”. In International Conference on Soft Computing and Data Mining , pp. 372-384. 2018
Kariyawasam, R. N. M. H. T. S. I. S. P. &. R. P., “Deep learning based screening and intervention of dyslexia, dysgraphia and dyscalculia”.. In 2019 14th Conference on Industrial and Information Systems , pp. 476-481. 2019.
Khan, R. U. C. J. L. A. &. B. O. Y., “Machine learning and Dyslexia: Diagnostic and classification system (DCS) for kids with learning disabilities”878 vb. International Journal of Engineering & Technology, 7(3.18), pp. 97-100. 2018.
| R. N. Keiron O Sea, “An Introduction to Convolutional Neural Networks,” IEEE, 2020. |
| A. S. u. Z. ,. A. S. Q. Asifullah khan, “A Survey of the Recent Architectures of Deep Convolutional Neural Networks,” Cornel University, 2019. |
| S. X. W. Y. M. Y. K. Ji, “convolutional neural networks for human actionrecognition. Pattern Analysis and Machine Intelligence,,” IEEE Transactions on 35(1),221–231 , 2017. |
| A. T. G. S. S. L. T. S. R. F.-F. L. Karpathy, “Large scale video classification with convolutional neural networks. In: Computer Visionand Pattern Recognition (CVPR),,” IEEE Conference on. pp. 1725–1732. IEEE, 2016. |
Know more about UniqueSubmission’s other writing services:
