| February 28, 2020 | 0 Comments

Usman Nazir, Usman Khalid Mian, Muhammad Usman Sohail, Murtaza Taj and Momin Uppal


The availability of high-resolution satellite imagery has enabled several new applications. One such application is identification of brick kilns for the elimination of modern day slavery which is also one of UN’s Sustainable Development Goals (SDG). This requires automated analysis of approximately 1,551,997 square kilometer area within ‘’Brick-Kiln-Belt’’ of South Asia. Although recent advancements in machine learning have achieved high accuracy for wide variety of applications, however problems involving such large-scale analysis particularly using high-resolution satellite imagery requires both accuracy as well as computational efficiency. We propose a coarse-to-fine strategy consisting of an inexpensive classifier and a detector which work in tandem to achieve high accuracy at low computational cost. To this extent we propose a two stage gated neural network architecture called Kiln-Net. At the first stage imagery is classified using ResNet-152 model which filters out over 99% of irrelevant data. At the second stage the remaining data is passed through a YOLOv3 based object detector to find the precise location of each brick kiln in the area. The framework is developed by training the network on the satellite imagery consisting of 14 different classes of South Asian region. The dataset, named as Asia14, consisting of 14000 images is also developed during the process. The proposed data includes Digital Globe RGB images of 14 different classes including brick kilns. Our proposed network architecture outperforms state-of-the-art architectures employed for the recognition of brick kilns and achieved 99.96% accuracy when tested on approximately 3300 square kilometer region (3,37,723 image patches) from 14 different cities of South Asia. To the best of our knowledge it is also 20x faster than the existing methods.

Asia14 Dataset: Asia14
Annotated Dataset: Additional Training Set
Semi-automated verification contributed annotations: Contributed Annotations
Evaluation Set (ROI is shaded)+ Contributed Annotations: Evaluation Set


Category: Journal Papers, Projects, Publications

Leave a Reply

Your email address will not be published. Required fields are marked *