As the field of industrial automation continues to evolve, the challenges of instance segmentation and object detection in computer vision and robotics remain significant. This project first aims to address these challenges by leveraging RGB, depth only, or RGB-D images to detect 2D objects. Specifically, such detection is intended for use in automated sorting and counting systems for commercial activities, such as fruit sorting in supermarkets. Compared to 2D images, 3D images give us a better representation of this world. A better representation of our living world can help automated systems understand the world more accurately. Then, depth cameras will be employed to capture 3D point clouds (3D features of objects), generating detailed features for the objects. To create a robust experimental dataset, 3D data pre-processing algorithms will be employed. Such a dataset will enable the improvement of the accuracy of automated systems in comprehending and processing the surrounding environment.