Computer Science > Computer Vision and Pattern Recognition
[Submitted on 9 Aug 2021]
Title:Towards artificially intelligent recycling Improving image processing for waste classification
View PDFAbstract:The ever-increasing amount of global refuse is overwhelming the waste and recycling management industries. The need for smart systems for environmental monitoring and the enhancement of recycling processes is thus greater than ever. Amongst these efforts lies IBM's Wastenet project which aims to improve recycling by using artificial intelligence for waste classification. The work reported in this paper builds on this project through the use of transfer learning and data augmentation techniques to ameliorate classification accuracy. Starting with a convolutional neural network (CNN), a systematic approach is followed for selecting appropriate splitting ratios and for tuning multiple training parameters including learning rate schedulers, layers freezing, batch sizes and loss functions, in the context of the given scenario which requires classification of waste into different recycling types. Results are compared and contrasted using 10-fold cross validation and demonstrate that the model developed achieves a 91.21% test accuracy. Subsequently, a range of data augmentation techniques are then incorporated into this work including flipping, rotation, shearing, zooming, and brightness control. Results show that these augmentation techniques further improve the test accuracy of the final model to 95.40%. Unlike other work reported in the field, this paper provides full details regarding the training of the model. Furthermore, the code for this work has been made open-source and we have demonstrated that the model can perform successful real-time classification of recycling waste items using a standard computer webcam.
Submission history
From: Ryan Grammenos Dr [view email][v1] Mon, 9 Aug 2021 21:41:48 UTC (1,381 KB)
Current browse context:
cs.CV
References & Citations
Bibliographic and Citation Tools
Bibliographic Explorer (What is the Explorer?)
Connected Papers (What is Connected Papers?)
Litmaps (What is Litmaps?)
scite Smart Citations (What are Smart Citations?)
Code, Data and Media Associated with this Article
alphaXiv (What is alphaXiv?)
CatalyzeX Code Finder for Papers (What is CatalyzeX?)
DagsHub (What is DagsHub?)
Gotit.pub (What is GotitPub?)
Hugging Face (What is Huggingface?)
Papers with Code (What is Papers with Code?)
ScienceCast (What is ScienceCast?)
Demos
Recommenders and Search Tools
Influence Flower (What are Influence Flowers?)
CORE Recommender (What is CORE?)
arXivLabs: experimental projects with community collaborators
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.