Application of Object Detection Models for the Detection of Kitchen Furniture - A Comparison

Object detection is being applied in an increasing number of areas. In this paper, the authors investigate the application of object detection in a use case for the kitchen industry. The main goal of the use case is to extract information from kitchen scenes that can be used for kitchen planning. The use case is located in a medium-sized company that has little experience with the application of deep learning models. Therefore, this paper proposes a methodology that ensures fast and reliable testing of different object detection models to identify a suitable model for the given use case. In the first step, a dataset with kitchen images is built. Further, augmentation methods are applied to the dataset, to increase the amount and variety of the data. For object detection, there is a variety of models that are freely available, and the question of which model is best for the use case cannot be answered easily. A selection of models (Faster R-CNN, SSD, and EfficentDet) from the TensorFlow Object Detection API will therefore be tested on the image dataset created. The achieved mean average precision (mAP) of the trained models will be used as a metric to determine the best model for the use case. The purpose of this work is to provide an approximate solution, that proves that object detection and the methodology work for the use case.