By Atharva Agrawal
Growing up in the Tiger Capital of India, Nagpur, a city surrounded by some of the country’s most eminent wildlife sanctuaries, including Pench National Park, Tadoba-Andhari, Kanha Tiger Reserve and Bandhavgarh National Park, I have always had a deep connection with untamed forests and quaint hamlets of rural India.
I have spent every summer and every birthday participating in game drive safaris, enabling me to closely understand and observe animal behaviour, including the protective partnership between chital and gray langurs.
My time in the wilderness and the conversations with locals and tour guides exposed me to a pressing issue: Human-Wildlife Conflict. This is the negative interaction between humans and wildlife, originating from competition for space and resources, resulting in harm, injury or death to both parties. I was dismayed when I learned about families in the Pilibhit region of UP sacrificing elderly members for compensation, widow villages in the Sundarbans, children escorted to school to avoid bear attacks, and communities stigmatized after tiger encounters. What struck me most was how quickly animals were labeled “man-eaters” and eliminated, while the underlying causes remained unaddressed.
Every time I glanced at the newspaper, I read about the dissension between humans and other species and environmental issues like rapid global warming, habitat encroachment and pollution. These issues were only going to amplify, and this is when I knew I had to do something!
As I began studying existing mitigation methods, I learned how ground patrolling was dangerous and resource-intensive, camera traps were reactive, and electric fencing was costly and risky. As a technophile, I saw an opportunity to intervene earlier.
In 2023, as AI tools became more accessible, I started experimenting with machine learning using Python. I worked with convolutional neural networks, TensorFlow Lite for edge deployment, and YOLOv8 for real-time object detection. A central question emerged: could a system listen to animals and their environment to reduce human–large mammal conflict? This question led to RAPTOR (Rapid-Alert and Patrol Tracking for Operational Rescue).
Offering a novel paradigm for wildlife conservation, RAPTOR combines acoustic monitoring and computer vision to detect animal distress calls and illegal human activity in real time. Built on Raspberry Pi hardware with integrated microphones and cameras, the system processes audio and video locally in remote environments.
Acoustic signals are converted into Mel-spectrograms and MFCC features, then classified using CNNs, while visual input identifies humans or suspicious movement in restricted areas. Outputs from both modalities are fused to reduce false positives. TensorFlow Lite enable slow-latency inference on limited hardware. RAPTOR can detect instances of poaching, conflict and intrusion and inform the authorities before it is too late. Moreover, it is designed for low-power and rugged deployment in forests.
After the development of the prototype, I filed for a provisional patent for the device. Additionally, under the mentorship of Dr Andrew Farnsworth from Cornell Lab of Ornithology, I authored a research paper
titled ‘An AI-Powered Acoustic Monitoring System for Anti-Poaching and Animal Distress Detection’. It includes findings on RAPTOR’s performance (95% accuracy in detecting intrusions with an alert speed under 2 seconds via wireless transmission) and efficiency compared to alternatives such as camera traps and on-foot patrolling.
I am currently in the process of pilot testing and deploying RAPTOR after receiving approvals from the Chief Conservator of Navegaon National Park in collaboration with an IIT Bombay Climate Tech start-up and the Wildlife Trust of India (WTI). My long term goal is to completely mitigate Human - Large Mammal conflict in Asia and Africa, and I aspire to become a conservationist and tech entrepreneur in the future.
