BanRakshak
Giving Forests a Voice Before They're Silenced
Created on 12th July 2025
•
BanRakshak
Giving Forests a Voice Before They're Silenced
The problem BanRakshak solves
The Problems BanRakshak Aims to Change
1. Invisibility of Illegal Deforestation
Current forest monitoring systems are reactive, delayed, or manually dependent. BanRakshak transforms this by enabling real-time detection of deforestation activity using sound, making illegal logging visible and actionable the moment it happens.
2. Lack of Immediate Alerts in Remote Forest Zones
Most forest crimes occur in off-grid, inaccessible regions where connectivity is limited. BanRakshak changes this by operating independently on solar energy and using GSM/GPS modules to send alerts without needing internet or human presence.
3. Human Resource Dependency for Patrolling
Manual forest patrols are limited, expensive, and ineffective at scale. BanRakshak reduces the burden on forest rangers by offering a distributed, automated early warning system that can monitor large regions 24/7.
4. Absence of Data-Driven Forest Protection
There is a lack of real-time, structured, and scalable data from forest ecosystems. BanRakshak introduces a new data pipeline through acoustic monitoring and machine learning, enabling data-driven decision making in conservation efforts.
5. Unchecked Biodiversity Threats
Illegal logging and poaching are major threats to wildlife and biodiversity. BanRakshak changes the current passive response model into a proactive one, protecting endangered species by detecting human intrusion before damage occurs.
6. Delayed Disaster Triggers (e.g., Landslide Risk)
Deforestation is directly linked to increased soil erosion and landslide risks. By preventing illegal tree-cutting activities in vulnerable zones, BanRakshak indirectly contributes to mitigating natural disaster triggers in rural and mountainous regions.
7. Misalignment with Global Climate Action
Despite climate goals, ground-level enforcement of forest protection is weak. BanRakshak brings a scalable, low-cost, and tech-driven solution that directly supports local implementation of climate action goals (SDG 13, 15, etc.).
8. Limited Use of Edge AI in Conservation
Conservation technology is often expensive or cloud-dependent. BanRakshak shifts this by showcasing how low-cost edge AI (on devices like Raspberry Pi) can solve high-impact environmental challenges without needing cloud infrastructure.
Summary
BanRakshak doesn’t just address a problem — it transforms how forests are protected.
It introduces a shift from manual to intelligent, delayed to real-time, and reactive to proactive forest conservation using edge-based AI and sound classification.
Challenges we ran into
Challenges Faced
-
Lack of Specific Audio Data
We lacked audio samples of axe cutting or tree chopping, which required us to manually crop relevant sounds from YouTube videos and collect additional samples from platforms like Freesound. -
Insufficient Test Dataset Context
Our test dataset lacked real-world forest context, compelling us to generate custom datasets or extract relevant segments from YouTube videos to simulate realistic scenarios. -
IoT Integration Complexity
Integrating the IoT device with the ML model and mapping audio files posed technical challenges. Fine-tuning the softmax threshold to distinguish between deforestation sounds and background forest noise was particularly difficult. -
Lack of Localized Data
Due to the absence of Nepal-specific acoustic datasets, we had to pivot our research toward globally available data. We relied on academic resources like the paper "Preventing Illegal Deforestation Using Acoustic Surveillance" to guide our approach. -
Low-Quality Microphone Issues
Our budget-friendly microphone introduced significant background noise, which degraded the quality of the captured audio and affected model accuracy. -
System Integration Difficulties
Running the frontend, backend, and IoT components in parallel created synchronization and performance challenges, especially during live testing and demos.
Tracks Applied (1)
Data Science / Machine Learning
Technologies used