Research methodology is the process that demonstrates how a project is carried out with a clear outline of the methods to be used. Its shows how the research goal would be realized through the systematic execution of the objectives. To accomplish the objectives of the research sequential neural network model is being developed, trained and evaluated using python programming language, tensorflow and kersa libraries and Visual studio code for the mobile application. Sequential Neural Network technique is considered most outstanding in the processing of sequential data of e-waste (Graves & Jaitly, 2014).
Data Collection
The dataset used for this research was collected from kaggle online dataset repository. The training data was obtainable which was mainly used in the classification of e-waste for recyclability status, which was classified as battery, computer, keyboard, mouse, printer, washing machine, PCB, player, microwave, mobile, television and speaker. There were about 3600 pictures with 300 images of battery, 300 images computers, 300 images keyboards, 300 images microwaves, 300 images mobiles, 300 images mouse, 300 image PCB, 300 images players, 300 images printers, 300 images speakers, 300 images washing machine and 300 images television (6) were processed, trained, tested and evaluated for performance. These categories were created in accordance with the images in the corresponding, independent folder.. Figure 2, below are images of some of the dataset that used for this research are classified as battery, computer, keyboard, mouse, printer, washing machine, pcb, player, microwave, mobile, television and speaker. There were about 3859 images and 3139 was used for training the model, 360 images each was used for validation and testing the model with 360 images of battery, 310 images computers, 330 images keyboards, 320 images microwaves, 330 images mobiles, 322 images mouse, 330 image PCB, 321 images players, 330 images printers, 306 images speakers, 324 images washing machine and 326 images television.
Table 1
Total Quantity of Pictures per Category
S/N | Categories | Training | Testing | Validation | Total |
1 | Battery | 250 | 30 | 30 | 310 |
2 | Computer | 250 | 30 | 30 | 310 |
3 | Keyboard | 270 | 30 | 30 | 330 |
4 | Mouse | 262 | 30 | 30 | 322 |
5 | Printer | 270 | 30 | 30 | 330 |
6 | Washing Machine | 264 | 30 | 30 | 324 |
7 | PCB | 270 | 30 | 30 | 330 |
8 | Player | 261 | 30 | 30 | 321 |
9 | Microwave | 260 | 30 | 30 | 320 |
10 | Mobile | 270 | 30 | 30 | 330 |
11 | Television | 266 | 30 | 30 | 326 |
12 | Speaker | 246 | 30 | 30 | 306 |
| Total | 3139 | 360 | 360 | 3859 |
The proposed Model
The study uses sequential neural network (SNN) architecture of deep learning algorithm to prepare four (4) mother-class, and each mother class contain three (3) categories of e-waste dependents of battery, computer, keyboard, mouse, printer, washing machine, PCB, player, microwave, mobile, television and speaker. Twelve convolutional neural networks are prepared into four blocks (A, B, C and D) each which contains three of the classes of the e-waste components one for each mother class, which will also prevent the recognition of a class that doesn't exist. For instance, if the mother class is computer, at that point we realize that the child class can't be a battery. At first, this model gets the image as input; at that point, the SNN creates the bounding box and the mother-class as an output. With that data, the real image is being cropped, which is one of CNN's input identified with the mother-class. After that, the output generated is the child class, which converged with the mother-class, and creates the formation of the final class. Finally, before the detection of the model, the output gets predicted with the bounding box. The mechanism of this model can be found in Fig. 3
A Sequential Neural Network
A Sequential Neural Network is a type of neural network architecture in Keras, a popular deep learning library. It's a linear stack of layers, where each layer is a fully connected neural network with an activation function. The output from one layer is used as input to the next layer, allowing the network to learn complex representations of the input data. The proposed architecture uses a sequential neural network (SNN CNN) followed by multiple specialized CNNs,.