Liquid Detection and Instance Segmentation based on Mask R-CNN in Industrial Environment

The goal of the paper is to present an efficient approach to detect and instantiate liquid spilled in the industrial and industrial-like environments. Motivation behind it is to enable mobile robots to automatically detect and collect samples of spilled liquids. Due to the lack of useful training data of spilled substances, a new dataset with RGB images and masks was gathered. A new application of the Mask-RCNN-based algorithm is proposed which has the functionalities of detecting the spilled liquid and segmenting the image.