Study on the CNN model optimization for household garbage classification based on machine learning

Xie, Wenzhuo; Li, Shiping; Xu, Wei; Deng, Hongkuan; Liao, Weihan; Duan, Xianbao

doi:10.3233/ais-220017

Cited by 5 publications

(2 citation statements)

References 25 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Transfer Learning [21] refers to the application of already learned knowledge or experience from one task to another to facilitate machine learning. The use of transfer learning can help to train an accurate and generalization capable model on a limited dataset [22].…”

Section: Transfer Learning Methodsmentioning

confidence: 99%

Garbage Classification Algorithm Based on Improved MobileNetV3

Tian,

Shi,

Luo

et al. 2024

IEEE Access

View full text Add to dashboard Cite

In recent years, the amount of household waste has increased sharply, and there is an urgent need to use intelligent waste classification equipment to assist in completing waste classification tasks. However, existing garbage classification algorithms have large parameter counts and poor real-time performance, which are not suitable for embedded garbage classification devices. Therefore, in order to achieve a lightweight and efficient classification model, this article uses a self-built garbage dataset and a pre-trained MobileNetV3 model in the PyTorch framework for garbage recognition and classification. We introduce the CBAM attention mechanism in the network model to enhance spatial feature perception. Utilize the Mish activation function to fully utilize the extracted depth image information. Using global average pooling instead of the fully connected layer of the original model reduces the number of model parameters while improving the recognition accuracy of the model. Finally, we propose an improved lightweight model called GMC-MobileNetV3. The experimental results show that the recognition accuracy of the improved MobileNetV3 model on self-built data set reaches 96.55%, which is 3.6% higher than that of the original model, the number of model parameters is 0.64M, the memory resource consumption is reduced by 56.6%, and the recognition time of a single garbage image is only 26.4ms. The network proposed in this article can achieve low consumption and high accuracy in garbage recognition and classification, providing reference for future academic research and engineering practice.

show abstract

Section: Transfer Learning Methodsmentioning

confidence: 99%

Garbage Classification Algorithm Based on Improved MobileNetV3

Tian,

Shi,

Luo

et al. 2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…For this reason, high-performance models trained on existing datasets were required so as not to reverse the cumulative knowledge for a new application. The TL was utilized to ensure the success of models on datasets consisting of fewer instances [22]. In other words, there are known TL architectures that contain pre-trained models, including beneficial knowledge to be used for a new problem.…”

Section: Transfer Learningmentioning

confidence: 99%

A Real-Time Nut-Type Classifier Application Using Transfer Learning

Özçevik

2023

Applied Sciences

View full text Add to dashboard Cite

Smart environments need artificial intelligence (AI) at the moment and will likely utilize AI in the foreseeable future. Shopping has recently been seen as an environment needing to be digitized, especially for payment processes of both packaged and unpackaged products. In particular, for unpackaged nuts, machine learning models are applied to newly collected dataset to identify the type. Furthermore, transfer learning (TL) has been identified as a promising method to diminish the time and effort for obtaining learning models for different classification problems. There are common TL architectures that can be used to transfer learned knowledge between different problem domains. In this study, TL architectures including ResNet, EfficientNet, Inception, and MobileNet were used to obtain a practical nut-type identifier application to satisfy the challenges of implementing a classifier for unpackaged products. In addition to the TL models, we trained a convolutional neural network (CNN) model on a dataset including 1250 images of 5 different nut types prepared from online-available and manually captured images. The models are evaluated according to a set of parameters including validation loss, validation accuracy, and F1-score. According to the evaluation results, TL models show a promising performance with 96% validation accuracy.

show abstract