Container Detection System Using CNN Based Object Detectors

The increase in free trade will also amplify the exchange of goods between countries and islands, especially in the seaports. The manual operation of the gantry-crane at the seaports has a risk due to human negligence. For that reason, automation is strongly needed in the loading and unloading of the container. Camera-based object detection is implemented to achieve this goal. In the recent years, CNN gives better accuracy in term of image classification, recognition and detection compared to the traditional method that use handcrafted features. In this paper, several advanced detection methods using CNN-based object detection, namely MobileNet, ResNet, and Faster RCNN are compared to detect and track the movement of containers. The results of the experiment show that the SSD ResNet and Faster RCNN methods are superior in terms of detection accuracy based on the Precision, Recall, Average Precision, and IoU values. MobileNet v3 excels in speed detection compared to the other methods.