top of page

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.

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.

result.png

Instance Segmentation Demo Videos

The following video shows instance segmentation of construction machinery with the model trained on the ACID dataset. 

Download

Please click the button below back to dataset download.

bottom of page