Application of Deep Learning in Garbage Classification

Shaoyou Ao

《International Core Journal of Engineering》 7卷9期 (2021/09) Pp. 262-266

https://doi.org/10.6919/ICJE.202109_7(9).0039

摘要

In terms of deep learning, after the entire process of research, Resnext101 32x8d WSL was finally used as the deep learning network model, sgd was the optimization function loss function using Cross Entropy Loss, the initial learning rate was set to 0.001, and the learning rate strategy of ReduceLROnPlateau was adopted, and adopted The Auto Augment algorithm is used as a data enhancement strategy. Finally, the accuracy of the deep learning model reached 0.9470, the precision reached 0.9212, the recall rate was 0.8420, and the F1 measurement value was 0.8799. Through intuitive observation and the test results of samples outside the data set, the model has good generalization ability and no overfitting. Since the Resnext101 32x8d WSL network has performed weakly supervised learning based on 940 million images, and its pre-training model is already relatively powerful, it has achieved good performance in actual application scenarios. After the data set is enhanced by AutoAugment, it is equal to Partial disturbances are added to enhance the robustness of the model.

關鍵字

Deep learning ； Resnet 50 ； AutoAugment ； Learning Strategy

參考文獻

Sharma N, Jain V, Mishra A. An analysis of convolutional neural networks for image classification[J]. Procedia computer science, 2018, 132: 377-384.

Gribbon K T, Bailey D G. A novel approach to real-time bilinear interpolation [C]// Proceedings. DELTA 2004. Second IEEE International Workshop on Electronic Design, Test and Applications. IEEE, 2004: 126-131.

Cubuk E D, Zoph B, Mane D, et al. Autoaugment: Learning augmentation strategies from data [C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019: 113-123.

Han D, Kim J, Kim J. Deep pyramidal residual networks[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2017: 5927-5935.

Nyoni T. Modeling and forecasting inflation in Kenya: Recent insights from ARIMA and GARCH analysis [J]. Dimorian Review, 2018, 5(6): 16-40.