Kiln-Net: A Gated Neural Network for Detection of Brick Kilns in South Asia

The availability of high-resolution satellite imagery has enabled several new applications such as identification of brick kilns for the elimination of modern-day slavery. This requires automated analysis of approximately 1 551 997 <inline-formula><tex-math notation="LaTeX">$\text{km}^2$</tex-math></inline-formula> area within the “Brick-Kiln-Belt” of South Asia. Although modern machine learning techniques have achieved high accuracy for a wide variety of applications, problems involving large-scale analysis 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. More specifically, we propose a two-stage gated neural network architecture called <italic>Kiln-Net</italic>. At the first stage, imagery is classified using the ResNet-152 model which filters out over <inline-formula><tex-math notation="LaTeX">$\text{99}\%$</tex-math></inline-formula> of irrelevant data. At the second stage, a YOLOv3-based object detector is applied to find the precise location of each brick kiln in the candidate regions. The dataset, named <italic>Asia14</italic>, consisting of <inline-formula><tex-math notation="LaTeX">$14\,000$</tex-math></inline-formula> Digital Globe RGB images and 14 categories is also developed to train the proposed kiln-net architecture. Our proposed network architecture is evaluated on approximately <inline-formula><tex-math notation="LaTeX">$\text{3{,}300}$</tex-math></inline-formula> km<inline-formula><tex-math notation="LaTeX">$^2$</tex-math></inline-formula> region (<inline-formula><tex-math notation="LaTeX">$337\,723$</tex-math></inline-formula> image patches) from 14 different cities in five different countries of South Asia. It outperforms state-of-the-art methods employed for the recognition of brick kilns and achieved an accuracy of <inline-formula><tex-math notation="LaTeX">$\text{99.96}\%$</tex-math></inline-formula> and average F1 score of 0.91. To the best of our knowledge, it is also <inline-formula><tex-math notation="LaTeX">$20\,$</tex-math></inline-formula>x faster than existing methods.

[1]  Dan G. Blumberg,et al.  New frontiers: Remote sensing in social science research , 1997 .

[2]  Sergey Ioffe,et al.  Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learning , 2016, AAAI.

[3]  Murtaza Taj,et al.  Tiny-Inception-ResNet-v2: Using Deep Learning for Eliminating Bonded Labors of Brick Kilns in South Asia , 2019, CVPR Workshops.

[4]  Andrew Zisserman,et al.  Very Deep Convolutional Networks for Large-Scale Image Recognition , 2014, ICLR.

[5]  Kaiming He,et al.  Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks , 2015, IEEE Transactions on Pattern Analysis and Machine Intelligence.

[6]  Ali Farhadi,et al.  YOLOv3: An Incremental Improvement , 2018, ArXiv.

[7]  Liang Chen,et al.  An Intensity-Space Domain CFAR Method for Ship Detection in HR SAR Images , 2017, IEEE Geoscience and Remote Sensing Letters.

[8]  Heekwan Lee,et al.  Integrated Environment Impact Assessment of Brick Kiln using Environmental Performance Scores , 2014 .

[9]  Ali Farhadi,et al.  You Only Look Once: Unified, Real-Time Object Detection , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[10]  Chao Tian,et al.  Dense Fusion Classmate Network for Land Cover Classification , 2018, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).

[11]  G. Foody,et al.  Slavery from Space: Demonstrating the role for satellite remote sensing to inform evidence-based action related to UN SDG number 8 , 2018, ISPRS Journal of Photogrammetry and Remote Sensing.

[12]  Jing Huang,et al.  DeepGlobe 2018: A Challenge to Parse the Earth through Satellite Images , 2018, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).

[13]  Qingshan Liu,et al.  Cascaded Recurrent Neural Networks for Hyperspectral Image Classification , 2017, IEEE Transactions on Geoscience and Remote Sensing.

[14]  Kenji Suzuki,et al.  A deep CNN based transfer learning method for false positive reduction , 2018, Multimedia Tools and Applications.

[15]  Ki-mook Kang,et al.  Ship Velocity Estimation From Ship Wakes Detected Using Convolutional Neural Networks , 2019, IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.

[16]  Bunkei Matsushita,et al.  Sensitivity of the Enhanced Vegetation Index (EVI) and Normalized Difference Vegetation Index (NDVI) to Topographic Effects: A Case Study in High-Density Cypress Forest , 2007, Sensors.

[17]  Shunxing Bao,et al.  Adversarial synthesis learning enables segmentation without target modality ground truth , 2017, 2018 IEEE 15th International Symposium on Biomedical Imaging (ISBI 2018).

[18]  Lei Guo,et al.  When Deep Learning Meets Metric Learning: Remote Sensing Image Scene Classification via Learning Discriminative CNNs , 2018, IEEE Transactions on Geoscience and Remote Sensing.

[19]  Qi Zhang,et al.  An Uncertainty Descriptor for Quantitative Measurement of the Uncertainty of Remote Sensing Images , 2019, Remote. Sens..

[20]  Jian Sun,et al.  Deep Residual Learning for Image Recognition , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[21]  Sang Michael Xie,et al.  Combining satellite imagery and machine learning to predict poverty , 2016, Science.

[22]  Ross B. Girshick,et al.  Mask R-CNN , 2017, 1703.06870.

[23]  Donald Geman,et al.  Coarse-to-Fine Face Detection , 2004, International Journal of Computer Vision.

[24]  Yu-Chiang Frank Wang,et al.  Deep Aggregation Net for Land Cover Classification , 2018, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).

[25]  Yunqian Ma,et al.  Imbalanced Learning: Foundations, Algorithms, and Applications , 2013 .

[26]  Dong ping Tian,et al.  A Review on Image Feature Extraction and Representation Techniques , 2013 .

[27]  Alexey A. Shvets,et al.  Feature Pyramid Network for Multi-class Land Segmentation , 2018, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).

[28]  Lelia Croitoru,et al.  Benefits and Costs of the Informal Sector: The Case of Brick Kilns in Bangladesh , 2012 .

[29]  Christopher E. Holden,et al.  Improved mapping of forest type using spectral-temporal Landsat features , 2018, Remote Sensing of Environment.

[30]  Giles M. Foody,et al.  Earth Observation and Machine Learning to Meet Sustainable Development Goal 8.7: Mapping Sites Associated with Slavery from Space , 2019, Remote. Sens..

[31]  Sujan Shrestha,et al.  A Comparative Study of Stack Emissions from Straight-Line and Zigzag Brick Kilns in Nepal , 2019, Atmosphere.

[32]  T. Landman,et al.  Globalization and Modern Slavery , 2019 .

[33]  Qingshan Liu,et al.  Learning Multiscale Deep Features for High-Resolution Satellite Image Scene Classification , 2018, IEEE Transactions on Geoscience and Remote Sensing.

[34]  Sergey Ioffe,et al.  Rethinking the Inception Architecture for Computer Vision , 2015, 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR).

[35]  Ciro Cattuto,et al.  Predicting City Poverty Using Satellite Imagery , 2019, CVPR Workshops.

[36]  Guigang Zhang,et al.  Deep Learning , 2016, Int. J. Semantic Comput..

[37]  Stefano Ermon,et al.  Deep Gaussian Process for Crop Yield Prediction Based on Remote Sensing Data , 2017, AAAI.

[38]  F. Giulio Tonolo,et al.  Building damage assessment scale tailored to remote sensing vertical imagery , 2018 .

[39]  Wei Liu,et al.  SSD: Single Shot MultiBox Detector , 2015, ECCV.

[40]  Elif Sertel,et al.  Comparative Research on Deep Learning Approaches for Airplane Detection from Very High-Resolution Satellite Images , 2020, Remote. Sens..

[41]  Sergey I. Nikolenko,et al.  Land Cover Classification from Satellite Imagery with U-Net and Lovász-Softmax Loss , 2018, 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops (CVPRW).

[42]  Franz J. Meyer,et al.  Automatic Ship Detection in Space-borne SAR Images , 2009 .

[43]  G. Meera Gandhi,et al.  Ndvi: Vegetation Change Detection Using Remote Sensing and Gis – A Case Study of Vellore District☆ , 2015 .

[44]  Stefano Ermon,et al.  Transfer Learning from Deep Features for Remote Sensing and Poverty Mapping , 2015, AAAI.

[45]  Chi-Wing Fu,et al.  H-DenseUNet: Hybrid Densely Connected UNet for Liver and Tumor Segmentation From CT Volumes , 2018, IEEE Transactions on Medical Imaging.