A Deep Learning-Based Approach to Garbage Detection in urban centers
The proposed method is an innovative way that address environmental concerns through the integration of artificial intelligence (AI) and image recognition technology. The application leverages the powerful Residual Network 101(ResNet101) model, trained on a dataset of over 3000 images, to accurately...
Saved in:
| Main Authors: | , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Tishk International University
2024-06-01
|
| Series: | Eurasian Journal of Science and Engineering |
| Subjects: | |
| Online Access: | https://eajse.tiu.edu.iq/index.php/eajse/article/view/25 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | The proposed method is an innovative way that address environmental concerns through the integration of artificial intelligence (AI) and image recognition technology. The application leverages the powerful Residual Network 101(ResNet101) model, trained on a dataset of over 3000 images, to accurately classify and analyze environmental issues depicted in user-submitted images. By utilizing the user's location data, the proposed method enables users to report environmental problems effectively, such as pollution, waste accumulation, deforestation, and more. For testing and practical applicability of the proposed approach, a mobile applicaton is developed to provide a user-friendly interface that allows users to capture images of environmental issues and submit them with relevant details. The images are then processed by the ResNet101 model, which employs deep learning techniques to classify and provide insights into the severity of the reported problems. The model achieved a low training loss of 0.0553 and a high training accuracy of 0.9809, indicating that learned well from the training data. The validation set's metrics, with a validation loss of 1.7773 and a validation accuracy of 0.6566, that show the model's performance on unseen data was relatively lower compared to the training set. However, the model's ability to generalize to new data is demonstrated by achieving a test loss of 0.8396 and a test accuracy of 0.8602 abstracts. |
|---|---|
| ISSN: | 2414-5629 2414-5602 |