Following the recent ban on plastic waste import by China, developed countries face challenges with a high amount of plastic waste. Plastic waste has been diverted to developing East-Asia countries like the Philippines, Vietnam, and Malaysia. The Malaysian government has taken strict action to send back over 3000 tons of contaminated plastic waste. This paper aims to establish mechanisms to detect the status of post-consumer PET bottles for recycling. A research-based and experimental design approach was adopted to develop mechanisms to detect PET bottle status to ensure high-quality bottles. A total of 1749 images were captured using a Raspberry Pi camera, belonging to four different classes: seal, cap, seal & cap, no seal, cap & content. Deep Learning technology (SqueezeNet) was used to train the PET bottle images. The trained model achieved 98% accuracy with correct bottle status recognized. The model was deployed on a Raspberry Pi to detect PET bottles in real-time. The model showed a delay of 0.018 to 0.022 seconds per prediction using Intel CPU in prediction performance. Whereas on Raspberry Pi, the prediction performance is 5 to 10 times slower than the Intel CPU, with a delay of 0.1 to 0.25 seconds per prediction.

How to Cite
WANINI MWANGI, MPAI MOKOENA, Hellen. Using Deep Learning to Detect Polyethylene Terephthalate (PET) Bottle Status for Recycling. Global Journal of Computer Science and Technology, [S.l.], nov. 2019. ISSN 0975-4172. Available at: <https://computerresearch.org/index.php/computer/article/view/1890>. Date accessed: 25 may 2020.