Machine Learning Sorting System Using Neural Network

Project Title

Machine Learning Sorting System Using Neural Network

Project Area of Specialization Electrical/Electronic EngineeringProject Summary

 Object sorting is a very common industrial application but at the same time it is a tiresome process as handling so many objects is a menial task which is not so promising in maintaining consistency and thereby arising quality issues. Object sorting, if done manually, is not only time consuming but also it seems to be an uphill task. To overcome these problem  we are creating a sorting system using neural  network. This sorting system  will be for waste management  because this existing waste segregation methods lack self-learning capability, are slow and inaccurate, hence requires constant replacement. They can work well in a small scale, but for large scale they are not very effective. This sorting system will separate the waste material into different classes for recycling. Classifying wastes into recycling categories using  Neural Networks can prove to be a very efficient methodology to process wastes.

Project Objectives

•With the help of Neural Network, a Machine Learning tool, high accuracy of waste segregation results can be achieved.

• In this project Convolutional Neural Network (CNN) is used.

• The neural network is trained for the various waste categories.

• By using neural network objects more than one characteristics can be distinguished.

• The trained network is integrated with a mechanical system that performs the physical segregation of waste, thereby avoiding human intervention.

Project Implementation Method

Dataset acquisition:

In order to train the neural network,we require huge amount of data only then the neural network can be trained efficiently. Higher the number of inputs, higher is the accuracy and vice-versa. 

Data Augmentation:

It adds value to basedataby adding information derived from internal and external sources within an enterprise.Basically what we do is for a same input image we do translation, rotation and scaling since it is not practically possible to take large number of images.

CNN Construction:

A convolutional neural network(CNN, orConvNet) is a class of deep,feed-forwardartificial neural networks, most commonly applied to analyzing visual imagery[5].It is a 3 layered network.

Training and Validation:

The model is trained on the training dataset using asupervised learningmethod. The training dataset often consist of pairs of an inputvectorandthe corresponding output vector, which is commonly denoted as thetarget. The current model is run with the training dataset and produces a result, which is then compared with thetarget, for each input vector in the training dataset. Based on the result of the comparison and the specific learning algorithm being used, the parameters of the model are adjusted. The model fitting can include bothvariableselectionand parameterestimation. Thefitted model is used to predict the responses for the observations in a second dataset called thevalidation dataset.The validation dataset provides an unbiased evaluation of a model fit on the training dataset

Testing:

Testing data is used to test the system. It is the set of data which is used to verify whether the system is producing the correct output after being trained or not.Testing data is used to measure the accuracy of the system.

Hardware Implementation:

The trained classifier is implemented on a raspberry pi which is coupled with a robotic actuator to pick and place the materials from the conveyor belt to the respective bins.

Benefits of the Project

• It is design to achieve segregation of waste thereby reducing human intervention in the handling of waste items.
•  The expansion in the waste categories database helps in increasing the accuracy rate when training the network for the classification purpose.
•  By using Neural Network in this project it can detect objects which is in noisy form.
•  By using Neural Network it has ability to train itself.

Technical Details of Final Deliverable

With the help of a Neural Network, a Machine Learning tool, high accuracy of waste segregation results can be achieved. By using this sorting mechanism objects with more than one characteristic be distinguished

Final Deliverable of the Project Hardware SystemCore Industry ManufacturingOther Industries Medical Core Technology Artificial Intelligence(AI)Other Technologies OthersSustainable Development Goals Good Health and Well-Being for People, Sustainable Cities and CommunitiesRequired Resources

Item Name	Type	No. of Units	Per Unit Cost (in Rs)	Total (in Rs)
			Total in (Rs)	79000
Raspberry pi	Equipment	2	15000	30000
Raspberry pi Camera module	Equipment	7	5000	35000
DC motor	Equipment	4	500	2000
servo motor	Equipment	4	500	2000
Conveyor belt	Miscellaneous	1	1000	1000
metal stand	Miscellaneous	4	1000	4000
basket	Miscellaneous	10	100	1000
other stationary	Miscellaneous	8	500	4000