Towards Scalable Identification of Brick Kilns from Satellite Imagery with Active Learning

Identifying illegally constructed brick kilns will help mitigate their environmental impact. Due to the sparse availability of Indian brick kailn data, we employed transfer learning and fine-tuned the models developed for Bangladesh data. Labeling brick kilns is time-consuming; thus, employing active learning we labeled data points where the model exhibited the highest uncertainty. With the trained model, we identified over 700 brick kilns in the Indo-Gangetic region and we are further expanding this approach for other regions in India.