The most popular applications of computer vision are surrounding object detection, object tracking, segmentation, action recognition, and anomaly detection. Although various effective AI models are developed and published as open source for each task, the effort to apply them to a specific need remains huge. This study proposed time-saving processes for developing computer vision systems by integrating state-of-the-art models into only one web-based system. We also present an object-level data augmentation approach: although the number of required data for training is reduced, the accuracy of models will increase. It is important for no-code systems where users may not have access to large datasets. From now on, computer vision systems can be developed with no line of code and require a small portion of time. This study was experimented with in reality and also validated on datasets Trashnet and AquaTrash, which achieved impressive results.