High School Student Develops Neural Network to Detect Gunshots for Forest Poaching Monitoring

Wildlife poaching remains a major threat in forests worldwide. Researchers use acoustic recorders to listen for gunshots, but detecting them amid rain, wind, and animal sounds is challenging. Naveen Jr, a 17-year-old high school student from San Diego, developed a compact neural network that can pick gunshots out of complex rainforest soundscapes with high recall and precision. By converting audio into spectrogram images and training a lightweight model, his approach aims to enable real-time monitoring in remote forests, helping rangers intercept poachers more efficiently while reducing dangerous encounters. The story also highlights the broader context of conservation, socioeconomic factors behind poaching, and the practical hurdles to field deployment.

Context: Poaching, Elephants, and Acoustic Monitoring

Wildlife poaching poses a severe threat to forest elephants and other species across Central and West Africa, as well as parts of Asia. In response, conservationists deploy sound recorders hidden high in canopies to monitor forest activity. Acoustic monitoring captures a wide range of biodiversity and human activity, offering an unbiased way to map gunfire events and track poaching patterns without always needing visual confirmation.

"Acoustic monitoring is really great at recording these soundscapes and getting this really amazing picture of biodiversity by eavesdropping on nature." - Daniela Hedwig

Limitations of Traditional Monitoring Approaches

Trail cameras and other standard monitoring tools have fundamental limitations. They offer narrow detection ranges, can be destroyed or stolen, and require proximity to hunting activity. Acoustic devices, by contrast, can be hidden away and cover vast areas, but distinguishing gunshots from natural sounds (branches breaking, rainfall, animal vocalizations) is difficult in noisy forests. Researchers emphasize that gunshots can be embedded in long recordings and may be faint at long distances, creating false positives that burden human reviewers.

"The detectors that were detecting the sounds of the gunshots, they either had too high of a false positive rate to be deployed in the field" - Naveen

Naveen Jr: A Breakthrough by a 17-Year-Old

Naveen Jr, a high school student from San Diego, decided to address poaching through acoustic data. He learned Python, studied the literature on gunshot detection, and built a neural network designed to be accurate yet lightweight enough to run in real time on field devices. He pursued a habitat-agnostic approach, aiming for a detector that could generalize from Belizean forests to African rainforests and beyond, avoiding overfitting to a single data set.

"I have always been interested in the natural world as far as I can remember, since like elementary school and then going through middle school and high school, and this whole project of building this neural network to detect poaching" - Naveen

Technical Approach: Spectrograms and Lightweight Models

To classify gunshots, Naveen transforms audio into spectrograms, visual representations of time and frequency with amplitude indicating loudness. He then uses a neural network built from the ground up to avoid overfitting to a single environment and to keep the model small enough to run on-device in real time. A key challenge in neural-network detectors is overfitting to habitat-specific acoustics, which reduces cross-site performance. Naveen overlays diverse sound examples from multiple recordings to teach the detector the characteristic rise and decay of gunshot spectrograms while distinguishing them from other sharp, non-gun sounds.

"In the few short years since we did our study, neural networks have really emerged as being a dominant approach to signal classification" - Naveen

Results: Cross-Habitat Generalization and Accuracy

Compared with a prior detector used by the elephant Listening Project at Cornell, which struggled with false positives, Naveen’s model achieved strong recall and precision across different data sets. When tested on a Cameroon data set, the prior template detector had recall around 87% with low precision; Naveen’s neural network achieved recall around 82% with precision around 0.87, and when trained on Belizean data, recall reached about 89% with precision around 0.93. The improvement in precision is particularly important in field deployments to keep rangers from chasing false alarms and to ensure prompt, reliable responses to genuine poaching events.

"the neural network he developed achieved a recall of 82% and a precision of 0.87" - Naveen

Implications for Conservation Patrols and Real-Time Monitoring

Higher accuracy in gunshot detection brings real-time acoustic monitoring closer to practical use in tropical forests. If detectors can reliably identify gunshots, anti-poaching patrols can respond immediately, potentially deterring poachers and enabling safer interventions. The approach also highlights broader benefits of acoustic sensing for biodiversity and habitat monitoring, providing an unbiased representation of human activity across large, inaccessible forest areas.

The researchers note several logistical hurdles to real-world deployment, including the need for power, storage, and reliable data transmission in remote areas. Solar solutions could help, but canopy cover and terrain can complicate power access. Nevertheless, the team envisions a future where lightweight, accurate detectors operate in challenging environments for extended periods, reducing the time lag between a poaching event and ranger response.

Broader Context: Socioeconomic Realities and Complex Solutions

The podcast also underscores that poaching is a complex, multi-faceted problem. While technology can boost detection and deterrence, socioeconomic drivers—poverty, food insecurity, and livelihoods—must be addressed to reduce the incentives for illegal hunting. The story emphasizes that many people entering national parks do so to feed their families, underlining the need for comprehensive conservation strategies that pair technology with community engagement and sustainable development efforts.

"It's important to acknowledge that there are many social and economic issues that contribute to poaching" - [Speaker]