Construction Instance Segementation
Instance segmentation is a computer vision task that involves identifying and delineating individual objects within an image. This algorithm can generate pixel-level masks of construction machinery, which provides additional mask information compared with object detection. Instance segmentation has great potential to enable real-time monitoring of construction sites in terms of productivity, safety, and resources.
Annotations
Instance segmentation annotation follows the same standard as object detection. Below are some examples of instance segmentation in ACID.
Accurate Mask labeling of the excavator and two dump trucks is included.
Accurate Mask labeling of the excavator and two dump trucks is included.
Accurate Mask labeling of the excavator and the dump trucks is included.
Contains accurate Mask labeling of the loader and two excavators.
Instance Segmentation Algorithm Analysis
We have divided the ACID dataset into a training set (80%) and a validation set (20%). Four deep learning instance segementation algorithms have been tested on the ACID dataset including Mask RCNN, YOLACT, SOLO, and SOLOv2.
Instance Segmentation Demo Videos
The following video shows instance segmentation of construction machinery with the model trained on the ACID dataset.