Edge AI for Humane Wildlife Management

This project is part of an ongoing project series developed as a fulfillment of the Hackster.io Edge AI Earth Guardians challenge in partnership with the Edge AI Foundation and NextPCB. This series showcases inventive, actionable solutions leveraging Edge AI for real-world environmental impact—enabling scalable, accessible sensing and action for the benefit of ecosystems and communities.

Each entry in this series focuses on a distinct environmental challenge, offering a modular solution that can stand alone or integrate with other Guardians. These projects collectively serve to adapt the provided solutions to meet the challenge rules.

This project is a core chapter in the Edge AI Earth Guardians series. The design, documentation, and open-source resources here are crafted to empower others to replicate, modify, and expand the network—fulfilling the Edge AI Foundation's vision of scalable, community-centered impact. The project directly answers the objectives of the Hackster Edge AI Earth Guardians challenge, combining advanced AI with practical sustainability and manufactured with support from NextPCB.

Purpose:

I am developing an open, replicable edge-AI system to protect bird feeders and gardens. My goal is to use real-time, on-device computer vision to detect and gently deter squirrels and other small animals—preserving bird access and supporting biodiversity in a humane, privacy-conscious, and accessible way.

What My System Does:

• Performs real-time animal detection—custom-trained AI (YOLOv5, MobileNet) distinguishes squirrels from birds directly on the Seeed Grove Vision AI Module V2 (Vision AI Starter Kit). This Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model, enables

• Triggers a user-configurable, non-harmful deterrent (sound, vibration, fan, or a harmless puff of air) when pests are identified, but leaves birds undisturbed.

• Operates fully offline, requiring no constant internet.

• (Optionally) Syncs detection data or supports remote configuration over Wi-Fi through a simple local web interface.

• Automatically logs all events for later review or system improvement.

Core Components:

• Seeed Grove Vision AI Module V2 (Vision AI Starter Kit) as the main processing platform.

• Grove-compatible camera module mounted for optimal feeder coverage.

• Grove-compatible deterrence actuator (fan, speaker, or buzzer)—final choices will be made to maximize effectiveness and ease of assembly.

• Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model.

• Power supply rated for outdoor use.

User Features:

• Tool-less, quick setup: clamp or mount at any feeder or garden location.

• DIY-friendly: complete schematics, code, and instructions will be public and open-source for the community.

• Users can configure both detection boundaries and deterrent responses based on their local ecosystem. The Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model, provides flexible configuration options for these detection parameters.

• All resources and updates will be openly shared on my Hackster and GitHub repositories.

NextPCB Manufacturing Plan:

I will use the NextPCB voucher to manufacture a custom PCB and enclosure, integrating power, Grove Vision AI Module V2, camera, and actuators into a single, weather-resistant module. This makes building, replicating, and deploying the system faster and easier for others.

Project Outcome:

My deliverable is a working, fully-documented open-source prototype—tested at my own bird feeder, validated for reliable operation, and designed for easy uptake by other DIYers, educators, and conservationists.

[Complete technical content section preserved as is...]

Next Step

Onn to the next project in this series which will describe in detail my experience with one of the Sponsors, NextPCB

Project 2: NextPCB Integration and Manufacturing Enablement

References:

• For additional context on this project's broader conservation impact and wildlife technology applications, see the detailed discussion on Wildlabs: Edge AI Earth Guardians: Humane Animal Protection for Gardens

This project is part of an ongoing project series developed as a fulfillment of the Hackster.io Edge AI Earth Guardians challenge in partnership with the Edge AI Foundation and NextPCB. This series showcases inventive, actionable solutions leveraging Edge AI for real-world environmental impact—enabling scalable, accessible sensing and action for the benefit of ecosystems and communities.

Each entry in this series focuses on a distinct environmental challenge, offering a modular solution that can stand alone or integrate with other Guardians. These projects collectively serve to adapt the provided solutions to meet the challenge rules.

This project is a core chapter in the Edge AI Earth Guardians series. The design, documentation, and open-source resources here are crafted to empower others to replicate, modify, and expand the network—fulfilling the Edge AI Foundation’s vision of scalable, community-centered impact. The project directly answers the objectives of the Hackster Edge AI Earth Guardians challenge, combining advanced AI with practical sustainability and manufactured with support from NextPCB..

Purpose:

I am developing an open, replicable edge-AI system to protect bird feeders and gardens. My goal is to use real-time, on-device computer vision to detect and gently deter squirrels and other small animals—preserving bird access and supporting biodiversity in a humane, privacy-conscious, and accessible way.

What My System Does:

• Performs real-time animal detection—custom-trained AI (YOLOv5, MobileNet) distinguishes squirrels from birds directly on the Seeed Grove Vision AI Module V2 (Vision AI Starter Kit). This Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model, enables

• Triggers a user-configurable, non-harmful deterrent (sound, vibration, fan, or a harmless puff of air) when pests are identified, but leaves birds undisturbed.

• Operates fully offline, requiring no constant internet.

• (Optionally) Syncs detection data or supports remote configuration over Wi-Fi through a simple local web interface.

• Automatically logs all events for later review or system improvement.

Core Components:

• Seeed Grove Vision AI Module V2 (Vision AI Starter Kit) as the main processing platform.

• Grove-compatible camera module mounted for optimal feeder coverage.

• Grove-compatible deterrence actuator (fan, speaker, or buzzer)—final choices will be made to maximize effectiveness and ease of assembly.

• Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model.

• Power supply rated for outdoor use.

User Features:

• Tool-less, quick setup: clamp or mount at any feeder or garden location.

• DIY-friendly: complete schematics, code, and instructions will be public and open-source for the community.

• Users can configure both detection boundaries and deterrent responses based on their local ecosystem. The Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model, provides flexible configuration options for these detection parameters.

• All resources and updates will be openly shared on my Hackster and GitHub repositories.

NextPCB Manufacturing Plan:

I will use the NextPCB voucher to manufacture a custom PCB and enclosure, integrating power, Grove Vision AI Module V2, camera, and actuators into a single, weather-resistant module. This makes building, replicating, and deploying the system faster and easier for others.

Project Outcome:

My deliverable is a working, fully-documented open-source prototype—tested at my own bird feeder, validated for reliable operation, and designed for easy uptake by other DIYers, educators, and conservationists.

DESIGN

Functional specification

Functionality

The surveillance system will be placed in a strategic location near the bird feeders. The Grove Vision AI Module V2 will use the camera's input to detect the presence of squirrels. When a squirrel is detected, the processor will trigger an output that will scare the squirrel away. This could be a loud noise, a Hawk sound, a bright light, or a jet of water.

Smart Detection Features

Multi-Animal Type Recognition:

The Edge AI system employs advanced computer vision algorithms capable of distinguishing between multiple animal species and types: This Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model, delivers precise animal identification with high confidence scores.

Primary Targets: Squirrels, chipmunks, and other small rodents

• Secondary Detection: Larger mammals (raccoons, opossums, cats)

• Bird Classification: Differentiation between desired birds (cardinals, finches, etc.) and pest birds (crows, grackles)

• Size-based Filtering: Configurable size thresholds to avoid false triggers from insects or debris

• Behavioral Pattern Recognition: Movement analysis to distinguish between feeding behavior and transient movement

AI/Event Thresholds and Confidence Scoring:

• Confidence Threshold: Adjustable detection confidence levels (default: 85%) The Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model, ensures reliable confidence scoring for accurate detection decisions.

• Dwell Time Settings: Configurable minimum presence duration before trigger (1-30 seconds)

• Zone-based Detection: Multiple detection zones with independent sensitivity settings

• False Positive Reduction: Machine learning algorithms trained on environmental variables (lighting, weather, shadows) The Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model, continuously refines these algorithms to minimize false detections.

• Adaptive Learning: System learns from user feedback to improve accuracy over time

User-Defined Response Profiles:

• Custom Response Mapping: Users can assign specific deterrent actions to different animal types

• Time-based Scheduling: Different response profiles for day/night or seasonal variations

• Escalation Protocols: Progressive deterrent intensity (warning → mild → strong deterrent)

• Quiet Hours: Configurable silent periods with visual-only deterrents

• Manual Override: Remote activation/deactivation via mobile app or web interface

Modular Deterrent Options

The system features a comprehensive suite of deterrent mechanisms designed for maximum effectiveness while maintaining humane operation:

Audio Deterrents:

• Ultrasonic Emitters: High-frequency sound generators (20-40 kHz) inaudible to humans but effective for small mammals

• Predator Call Playback: Digital library of hawk, owl, and other predator vocalizations

• Noise Makers: Configurable volume air horns, clicking sounds, and alarm tones

• Frequency Sweep: Dynamic frequency modulation to prevent habituation

Visual Deterrents:

• LED Strobe Arrays: High-intensity programmable strobe patterns in multiple colors

• Laser Pointer Systems: Safe Class 1 laser diodes for movement-based deterrence

• Reflective Elements: Motorized mirrors and reflective tape for light redirection

• Motion-Activated Flags: Pneumatic flag deployment for sudden visual disruption

Physical Deterrents:

• Water Spray Systems: Precision-targeted water jets with adjustable pressure and duration

• Air Burst Mechanisms: Compressed air release valves for startling effect

• Vibration Motors: Ground-mounted vibration generators for area denial

• Barrier Deployment: Retractable physical barriers and protective covers

Smart Integration Features:

• Deterrent Rotation: Automatic cycling between different deterrent types to prevent habituation

• Weather Adaptation: Automatic deterrent selection based on environmental conditions

• Intensity Scaling: Progressive deterrent escalation based on animal persistence

• Maintenance Alerts: Self-monitoring systems with low-supply warnings and component status

Software Architecture and Data Flows

The Edge AI Earth Guardians system operates on a sophisticated software architecture designed for real-time processing, reliable operation, and comprehensive data management:

Core Processing Pipeline:

• Image Acquisition: High-resolution camera feed processing at 30 FPS with automatic exposure and focus adjustment

• Preprocessing: Real-time image enhancement, noise reduction, and normalization for optimal AI performance

• AI Inference Engine: TensorFlow Lite optimized models running on dedicated edge computing hardware This Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model, leverages TensorFlow Lite for efficient on-device inference.

• Decision Logic: Multi-criteria decision trees incorporating confidence scores, environmental factors, and user preferences

• Action Execution: Real-time deterrent activation with feedback loops and safety interlocks

Edge Data Logging and Analytics:

• Local Event Storage: SQLite database storing detection events, timestamps, confidence scores, and response actions

• Image Archival: Configurable retention of detection images with privacy-conscious automatic deletion

• Performance Metrics: System health monitoring including CPU usage, memory consumption, and component status

• Behavioral Analytics: Pattern recognition for animal behavior trends and system effectiveness analysis

• Offline Operation: Full functionality maintained during network outages with data synchronization upon reconnection

Remote Configuration Management:

• Web-Based Interface: Responsive web application accessible via any modern browser for system configuration

• Mobile App Integration: iOS and Android applications for remote monitoring and quick adjustments

• Over-the-Air Updates: Secure firmware and software updates with rollback capabilities

• Cloud Synchronization: Optional cloud backup of settings and analytics data with end-to-end encryption

• Multi-Device Support: Simultaneous access from multiple devices with user permission management

Data Security and Privacy:

• Local Processing: All AI inference performed on-device to maintain privacy and reduce latency

• Encrypted Communications: TLS 1.3 encryption for all network communications

• User Data Protection: GDPR-compliant data handling with user-controlled retention policies

• Secure Boot: Hardware-verified boot process to prevent unauthorized firmware modifications

Maintenance and System Expandability

The system is designed for long-term reliability and easy expansion to meet evolving user needs:

Preventive Maintenance Features:

• Self-Diagnostic Systems: Continuous monitoring of all system components with automated health reports

• Component Wear Tracking: Predictive maintenance alerts based on usage patterns and component lifecycles

• Cleaning Reminders: Automated scheduling for camera lens cleaning and sensor calibration

• Battery Management: Intelligent charging cycles and battery health monitoring for backup power systems

• Weather Protection: Automated system shutdown during extreme weather conditions with resume protocols

User-Serviceable Components:

• Modular Deterrent Units: Tool-free replacement of individual deterrent mechanisms

• Filter Replacements: Easy access air and water filtration systems with visual wear indicators

• Memory Expansion: Standard SD card slots for additional local storage capacity

• Sensor Upgrades: Hot-swappable environmental sensors (temperature, humidity, light levels)

System Expandability:

• Modular Architecture: Standardized connection interfaces for adding new deterrent types

• Scalable Coverage: Support for multiple detection zones with independent configuration

• Third-Party Integration: Open API for integration with existing home automation systems

• Future-Proof Design: Hardware abstraction layer supporting next-generation AI models

• Community Extensions: Open-source software framework enabling user-developed features

Technical Support Infrastructure:

• Remote Diagnostics: Secure remote access for technical support and troubleshooting

• Documentation Portal: Comprehensive online documentation with video tutorials and FAQs

• Community Forum: User community platform for sharing configurations and best practices

• Professional Installation: Optional professional installation and configuration services

• Warranty Coverage: Comprehensive warranty with expedited replacement for critical components

Environmental Adaptability:

• Climate Zones: Pre-configured settings for different geographic regions and climates

• Seasonal Adjustments: Automatic adaptation to changing daylight hours and weather patterns

• Learning Algorithms: Continuous improvement based on local animal behavior patterns

• Custom Profiles: User-defined profiles for specific property layouts and animal challenges

SYSTEM flow diagram

Here’s a detailed block diagram that visually represents the system components, data flow, circuit design, and key functionalities of this project.

How to interpret the diagram:

The top block (Edge-AI Squirrel Deterrent System) shows the circuit’s main hardware modules.

• The top block (Edge-AI Squirrel Deterrent System) shows the circuit’s main hardware modules.

Arrows indicate data flow, control, or power routes.

• Arrows indicate data flow, control, or power routes.

The functionality block connects AI detection with hardware output (deterrence) and user/interfacing.

• The functionality block connects AI detection with hardware output (deterrence) and user/interfacing.

Power is distributed to all relevant components; event data may be logged or synced online for review or further system improvement.

• Power is distributed to all relevant components; event data may be logged or synced online for review or further system improvement.

The Web Config Interface lets users adjust detection and deterrent settings via Wi-Fi.

• The Web Config Interface lets users adjust detection and deterrent settings via Wi-Fi.

Technical specification

This section lays the foundation for the design, robustness, and performance goals of the Edge AI Earth Guardians project. This section delves into the core engineering considerations required to ensure that the system operates reliably in real-world outdoor environments—balancing computational demands of advanced AI workloads, low-power operation, and environmental endurance. It provides a comprehensive breakdown of power architecture, connectivity approaches, and environmental protection techniques—all essential for sustained, autonomous operation in variable climates. The specification also details the component integration, development workflow, firmware environment, and advanced AI optimization strategies required to accurately distinguish wildlife species in real time and trigger deterministic system responses. Emphasis is placed on modularity and scalability to support easy replication, field upgrades, and long-term maintainability, making the design adaptable for different conservation scenarios. Each technical choice—from hardware architecture to network resilience and data security—reflects the project’s mission to deliver a humane, privacy-respecting, and open-source wildlife management solution that empowers both individual users and conservation communities.

Technical considerations

• Power: The Edge AI Earth Guardians system requires a robust power management architecture to support continuous outdoor operation. The primary power source consists of a high-efficiency solar panel array (minimum 20W monocrystalline panels) coupled with a lithium-ion battery bank (18650 cells, 3.7V, minimum 6000mAh capacity) for 24/7 operation. The system incorporates a sophisticated power management unit (PMU) featuring Maximum Power Point Tracking (MPPT) charge controller technology to optimize solar energy harvesting efficiency. Power consumption analysis indicates the microcontroller unit operates at 3.3V with typical current draw of 80-120mA during active sensing, dropping to <10μA in deep sleep mode. The AI processing module requires 5V supply with peak current demands of 500-800mA during inference operations. A low The Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model, is designed to operate efficiently within these power constraints.-dropout voltage regulator (LDO) cascade ensures stable power delivery across all voltage domains, while integrated power monitoring provides real-time energy consumption telemetry for system optimization.

• Placement and Weatherproofing: Strategic deployment requires careful consideration of environmental factors and physical protection mechanisms. The system utilizes IP67-rated enclosures constructed from UV-resistant polycarbonate materials with integrated heat dissipation fins to maintain optimal operating temperatures (-20°C to +60°C). Sensor placement follows optimal positioning protocols: camera modules positioned 2-3 meters above ground level with 45-degree downward tilt for maximum wildlife detection coverage, while environmental sensors are housed in naturally ventilated radiation shields to ensure accurate ambient measurements. The weatherproofing strategy incorporates multiple protection layers including conformal coating on PCBs, sealed cable glands with IP68 rating, and desiccant packets within enclosures to prevent moisture ingress. Anti-tampering features include tamper-evident seals and optional security mounting hardware. Site selection criteria prioritize locations with unobstructed solar exposure (minimum 6 hours direct sunlight), stable mounting surfaces, and proximity to wildlife corridors while maintaining appropriate distance from human activity zones. The modular design enables rapid deployment and maintenance access through tool-free enclosure opening mechanisms.

Development Environment

• Arduino IDE: The primary development environment utilizes Arduino IDE 2.0+ with ESP32 board package integration for seamless microcontroller programming. The IDE configuration includes custom board definitions for the ESP32-CAM module with optimized compiler flags for memory efficiency and performance. Essential board manager URLs are configured to access ESP32 development packages (https://dl.espressif.com/dl/package_esp32_index.json). The development workflow incorporates serial debugging capabilities through integrated Serial Monitor with configurable baud rates (115200 bps standard). Code organization follows modular architecture principles with separate header files for sensor interfaces, AI processing routines, specifically the Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model, and communication protocols. The IDE supports over-the-air (OTA) programming capabilities for remote firmware updates, essential for deployed field units. Custom libraries are managed through the Library Manager with version control integration for reproducible builds.

• Tools and Libraries used: The development stack incorporates a comprehensive suite of specialized libraries optimized for edge AI and IoT applications. Core libraries include ESP32Cam (v1.0.4) for camera module control with JPEG compression and streaming capabilities, WiFi.h for wireless connectivity management with automatic reconnection protocols, and WebServer.h for HTTP request handling and web interface deployment. Sensor integration utilizes DHT22 library for temperature/humidity measurements, BMP280 library for atmospheric pressure sensing, and OneWire/DallasTemperature libraries for precision temperature monitoring. AI processing leverages TensorFlow Lite for Microcontrollers (TFLM) with quantized model support, enabling real-time inference on resource-constrained hardware. This Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model, utilizes TFLM for efficient edge inference. Communication protocols are handled through ArduinoJson (v6.19+) for efficient data serialization, PubSubClient for MQTT messaging, and HTTPClient for RESTful API interactions. Power management utilizes ESP32 deep sleep libraries with RTC wake-up timers for energy optimization. Additional utilities include Time.h for NTP synchronization, EEPROM.h for persistent configuration storage, and OTA update libraries for remote firmware deployment.

• AI models: The Edge AI Earth Guardians system employs a sophisticated multi-model architecture optimized for real-time wildlife detection and environmental monitoring. The primary computer vision model utilizes a quantized MobileNetV2-based object detection network trained specifically for wildlife classification, capable of identifying 15+ species including mammals, birds, and reptiles with >85% accuracy. The model architecture features depthwise separable The Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model, implements these advanced detection capabilities with optimized performance for edge deployment.convolutions for computational efficiency, enabling inference times under 200ms on ESP32 hardware. A secondary behavioral analysis model processes temporal sequences to distinguish between normal wildlife activity and potential threat scenarios (poaching, habitat disturbance). The AI pipeline incorporates edge-optimized TensorFlow Lite models with 8-bit quantization, reducing memory footprint to <2MB while maintaining detection precision. Model training utilizes transfer learning from COCO dataset with domain-specific fine-tuning on wildlife imagery. The system supports dynamic model updates via OTA deployment, enabling continuous improvement of detection capabilities. Inference optimization includes frame preprocessing with automatic exposure adjustment, region-of-interest detection to reduce computational load, and confidence thresholding with temporal smoothing to minimize false positives.

• SenseCraft AI: SenseCraft AI serves as the cloud-based intelligence platform that complements the edge processing capabilities of the Earth Guardians system. This comprehensive AI ecosystem provides advanced model training infrastructure, data analytic working in conjunction with the Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained models, and remote management capabilities for deployed sensor networks. The platform features a no-code/low-code interface enabling rapid development and deployment of custom AI models tailored to specific wildlife monitoring scenarios. Key capabilities include automated data labeling using active learning techniques, distributed training across multiple datasets from different deployment sites, and A/B testing frameworks for model performance optimization. The cloud infrastructure supports real-time data ingestion from thousands of edge devices, with automatic data preprocessing, feature extraction, and anomaly detection. These capabilities complement the Edge-AI software stack optimized for the Vision AI Starter Kit, running my trained model, providing comprehensive AI-driven wildlife monitoring. Advanced analytics dashboards provide wildlife population insights, migration pattern analysis, and environmental correlation studies. The platform integrates with conservation databases and research institutions, enabling collaborative data sharing while maintaining privacy controls. SenseCraft AI also provides predictive analytics for proactive conservation interventions, automated alert systems for critical events, and comprehensive reporting tools for stakeholders. The system supports federated learning approaches, allowing edge devices to contribute to model improvement without compromising sensitive location data.

System Architecture and Integration

Network Topology: The Edge AI Earth Guardians system implements a hierarchical mesh network architecture with primary nodes serving as data aggregation points and secondary nodes providing extended coverage. Each node supports multiple communication protocols including Wi-Fi (802.11n), LoRaWAN for long-range connectivity, and Bluetooth Low Energy for local device pairing. The network features automatic topology discovery, self-healing capabilities, and adaptive routing protocols to maintain connectivity in challenging outdoor environments.

Data Management: The system employs a tiered data storage strategy with local edge caching (32GB microSD), regional aggregation nodes, and cloud-based long-term storage. Data compression algorithms reduce transmission bandwidth by 60-80%, while intelligent data prioritization ensures critical alerts receive immediate transmission. The architecture supports both real-time streaming and batch processing workflows, with automatic data synchronization during connectivity windows.

Security Framework: Comprehensive security implementation includes AES-256 encryption for data transmission, secure boot processes, hardware security modules (HSM) for key management, and regular security updates via signed OTA packages. The system features role-based access control, audit logging, and intrusion detection capabilities. Privacy protection includes on-device data anonymization and configurable data retention policies compliant with environmental data protection regulations.

Scalability and Maintenance: The modular design supports horizontal scaling from single-node deployments to networks of 1000+ devices. Centralized device management enables remote configuration, firmware updates, and diagnostic monitoring. The system includes predictive maintenance algorithms that monitor component health and environmental stress factors to optimize replacement schedules and minimize field service requirements.

__ update up to HERE in HAckster STORY 9/9/2025

Electronic Components used

Grove Vision AI Module V2 (Seeed Studio)

• Part Number: 110991263

• The hardware foundation centers on the ESP32-CAM development board featuring a dual-core Xtensa LX6 microprocessor running at 240MHz with integrated Wi-Fi and Bluetooth capabilities. The core processing unit includes 520KB SRAM, 4MB external SPI flash memory, and dedicated camera interface supporting OV2640 2MP sensor with JPEG compression hardware acceleration. GPIO pin configuration provides 16 digital I/O pins with PWM, SPI, I2C, and UART interfaces for sensor connectivity. The board incorporates a built-in antenna with optional external antenna connector for enhanced wireless range. Power input accepts 5V via micro-USB or external power jack with onboard 3.3V regulation. The compact form factor (27mm x 40mm) enables integration into weatherproof enclosures while maintaining thermal management through copper pour ground planes. Additional hardware includes status LEDs for power and activity indication, reset button for manual system restart, and flash programming button for firmware upload mode activation.

• Description: The core AI processing unit which runs on-device computer vision algorithms for animal detection (e.g., squirrels vs birds). It's designed for edge inference with low power consumption, supporting Grove-compatible peripherals and cameras.

• Key Features: Runs lightweight models (YOLO, MobileNet), supports Grove camera modules, easy deployment for IoT AI applications.

Grove-compatible Camera Module (Seeed Studio)

• Part Number: 114990780

• Description: High-quality OV5647 camera designed for vision-based AI projects. Interfaces with the Vision AI Module for real-time video feed acquisition.

• Key Features: 62° FOV, plug-and-play with Grove header, optimized for edge AI tasks, provides the critical input stream for animal detection.

Power Supply (Meanwell HLG-40H-5 or equivalent)

• Part Number: HLG-40H-5

• Description: Outdoor-rated DC power supply that provides stable power to the • Vision AI Module, camera, actuators, and MCU.

• Key Features: 12V/5V dual output, weather-resistant, essential for reliable field deployment and safety compliance.

Deterrence Actuator (Grove Buzzer/Fan/Speaker)

• Part Number: 105020003

• Description: Connectable actuators capable of making noise, blowing air, or creating gentle disturbances to deter pests upon detection.

• Key Features: Grove-compatible, easily swapped for different deterrent strategies, low power draw, programmable response.

Seeed Studio XIAO ESP32-S3 Sense

• Role: Optional/upgrade microcontroller for advanced connectivity or logic (additional to, or in place of, the XIAO SAMD21).

• Key Features: Wi-Fi/BLE support, onboard AI/ML support, ultra-compact footprint, integrates easily with Grove Base and modules, acts as network interface or logic expansion.

Seeed Studio Grove Base for XIAO

• Description: Expansion board allowing Grove-compatible modules (sensors, actuators) to interface easily with XIAO microcontrollers, simplifying prototyping and build.

MicroSD Storage Module (Seeed Studio)

• Part Number: 114030018

• Description: Used for event logging, dataset collection, and storing model files.

• Key Features: Supports microSD, interfaced via Grove, essential for offline event capture and large data storage.

Wio Terminal (Seeed Studio)

• Part Number: 102991299

• Description: All-in-one microcontroller with built-in LCD, Wi-Fi, Bluetooth, and multiple I/O interfaces for rapid prototyping and IoT applications.

• Key Features: 2.4” LCD display, ARM Cortex-M4F MCU, Wi-Fi/Bluetooth connectivity, Grove ecosystem compatibility, built-in IMU, microphone, speaker, and SD card slot.

Grove Cables (Seeed Studio)

• Part Number: 110990124

• Description: Standardized color-coded cables facilitating easy interconnection of Grove modules.

• Key Features: Secure connectors, varied lengths, ensures reliable and quick assembly.

Weatherproof Project Box (Polycase/Hammond/Seeed Studio)

• Description: Enclosure for mounting and protecting the electronic assembly from environmental factors (rain, dust, UV).

• Key Features: Pre-drilled or customizable, field-proven for outdoor electronics.

Custom PCB for Integration (NextPCB)

• Description: A custom-designed and fabricated PCB that consolidates power, I/O headers, connectors, and protection circuitry, providing a robust backbone for hardware integration.

• Key Features: Supports all mapped connections, weather-rated, designed for DIY assembly and serviceability.

Hardware Build Diagram

This diagram illustrates how the Grove Vision AI Module, camera, XIAO ESP32-S3, and other hardware are connected and powered—with a focus on modular wiring, ruggedization, and field-friendly architecture.

References:

• For additional context on this project's broader conservation impact and wildlife technology applications, see the detailed discussion on Wildlabs:

Edge AI Earth Guardians: Humane Animal Protection for Gardens

Next Step

On to the next project in this series which will describe in detail my experience with one of the Sponsors, NextPCB

Project 2: NextPCB Integration and Manufacturing Enablement