Chasing Pelican based Deep Learning for Multiple Object Detection from Single Input Trash Image

Waste management is essential for developing a smart city to enhance the population's living standards. Harmonious and healthy living environments are achieved through waste sorting and recycling techniques. Waste object detection is crucial for waste sorting and recycling. Several methods were devised for the waste object detection in the image, but the inaccurate detection and computation overhead limit the model's performance. Hence, optimized deep learning named Chasing Pelican-based Adaptive Mask Region-Based Convolutional Neural Network (RCNN) is proposed for waste object detection to enhance detection accuracy with minimal computation overhead. The adaptive learning concept is incorporated in Mask RCNN to improve the accuracy of detection. Besides, optimal weight adjustment of Adaptive Mask RCNN using the Chasing Pelican algorithm minimizes the information loss during the data learning phase. The inclusion of the adaptation concept helps the model to better adapt to the features of waste objects in different environmental situations, resulting in more accurate detection results. Furthermore, the optimal weight adjustment of Adaptive Mask RCNN using the Chasing Pelican algorithm plays a crucial role in minimizing information loss during the data learning phase. This optimization ensures that the model efficiently learns and retains essential characteristics and patterns from the data, resulting in better detection performance. Thus, the proposed method accomplished enhanced accuracy in detecting the trash object based on F1-Score, recall, precision, accuracy, mAP, Mean Absolute Percentage Error (MAPE) and mean square error (MSE) obtained 96%, 99.75%, 99.75%, 99.875%, 99.75%, 0.3% and 0.4% respectively.