New Delhi: Scholars at Oxford University and the Lahore University of Management Sciences (LUMS) have leveraged AI to detect brick kilns from satellite imagery at unprecedented speeds.
This development could provide a fillip to authorities in taking action against illegal kilns, which not only pollute the air and degrade the soil but are also sites of exploitation.
There are over 55,000 brick kilns across a belt that stretches through Pakistan, India, Nepal and Bangladesh. Many of these are unregulated, but the region’s resource-strained law enforcement is unable to police these kilns through on-the-ground methods alone, a press release by Oxford University said.
Using aerial imagery is one way to expedite this process, but current technology means that doing so is either inaccurate or too expensive no matter what method we use.

A brick kiln and its chimney in Assam. Photo: Wikimedia Commons/অজয় দাস. CC BY-SA 4.0.
But that won’t be the case any longer, say Dr Sara Khalid, an associate professor at Oxford, and Dr Usman Nazir, a research fellow at LUMS, thanks to a new deep learning-based model that they’ve collaborated to build.
Deep learning is a subset of AI that uses artificial neural networks to mimic how the human brain learns about things.
“In the first stage of our model, the [aerial] images are first filtered at low-resolution by an algorithm that identifies potential brick kilns based on their spectral properties,” Dr Nazir said. “This means that only a small subset proceeds to the second stage of fine-grained localisation, where objects are detected at high-resolution using a deep learning-based model.”

Aerial view of a brick kiln. Photo: https://doi.org/10.48550/arXiv.2303.11654.
“Having these two stages reduces the required computational power considerably, speeding up the process,” he added.
And the potential speeds this process can now take on are unprecedented. While it would normally take anywhere from 92 days to five years to process imagery data from the entire brick kiln-belt, using the new model would take it just four days.
India produces the world’s second-largest number of bricks annually.
Read previous coverage by The Wire about India’s lakhs of brick kilns, where workers live in squalid conditions, political nexuses override environmental safeguards and day-to-day functioning is structured around exploiting the illiteracy of workers.