<![CDATA[Plastic bottle waste must be separated based on shape and size to facilitate recycling. Sorting plastic bottles can use object detection technology to facilitate classification using images. Image distance capture affects the classification of bottle waste because large bottles will look small when seen from a distance and vice versa. This study aims to create a plastic bottle detection and classification model using the YOLOv8 algorithm with the same bottle shape but different sizes and measure the effect of image distance on the model. Bottles consist of three sizes: large bottles measuring 1500 ml, medium bottles measuring 600 ml, and small sizes 330 ml. Pictures for the bottle image dataset were shot between 80 and 100 centimeters away. Robotoflow was used to produce the dataset. Model performance evaluation used Mean Average Precision, and model testing used a confusion matrix. The test results for the same model with an image capture distance had an accuracy value of 100%. Testing of 80 cm distance images applied to the 100 cm model had an accuracy of 67%. Testing for 100 cm distance images applied to the 80 cm model was still quite good, with an accuracy of 91.6%. The results obtained show that the image distance affects the results of the model that has been built, so use an image that matches the distance applied to the model.]]>
Copyrights © 2024
