bugcam 0.1.5

detect and classify insects on a raspberry pi

bugcam - Raspberry Pi Insect Detection

CLI for running insect detection on Raspberry Pi with Hailo AI HAT+.

Hardware Requirements

• Raspberry Pi 4/5
• Raspberry Pi Camera Module 3
• Raspberry Pi AI HAT+ (Hailo8/Hailo8L)
• Active cooler recommended

Installation

1. System Setup

Update system packages:

sudo apt update && sudo apt upgrade -y

Install Hailo hardware drivers (required for AI HAT+):

sudo apt install hailo-all

Verify camera is detected:

rpicam-hello

If this fails, enable the camera interface in raspi-config:

sudo raspi-config
# Navigate to: Interface Options > Camera > Enable
# Reboot when prompted

2. Install bugcam

Option A: System-wide install (simplest)

pip install bugcam

Option B: Isolated install with pipx (recommended)

# Install pipx
sudo apt install pipx

# Add pipx binaries to PATH
pipx ensurepath

# IMPORTANT: Close and reopen your terminal for PATH changes to take effect
# Then install bugcam
pipx install bugcam

3. Download Detection Model

# Download the small model (11MB, faster)
bugcam models download yolov8s

# OR download the medium model (31MB, more accurate)
bugcam models download yolov8m

# List installed models
bugcam models list

Models are downloaded from S3 and cached in ~/.cache/bugcam/models/.

4. Run Detection

# Preview camera with detection overlay
bugcam preview

# Start continuous detection
bugcam detect start

# Save detections to file
bugcam detect start --output detections.jsonl

# Enable autostart on boot
bugcam autostart enable

CLI Reference

Models

Manage and inspect detection models.

Download a model:

bugcam models download <model_name>

Available models: yolov8s (11MB, faster), yolov8m (31MB, more accurate)

List installed models:

bugcam models list

Show model details:

bugcam models info <model_name>

Preview

Run live camera preview with detection overlay.

# Basic preview
bugcam preview

# Preview with specific model
bugcam preview --model yolov8m

Options:

• --model <name> - Specify which model to use (default: yolov8m)

Detection

Run continuous detection and save results.

# Start detection
bugcam detect start

# Save detections to file
bugcam detect start --output detections.jsonl

# Run for specific duration (in minutes)
bugcam detect start --duration 30

# Quiet mode (suppress console output)
bugcam detect start --quiet

# Combined options
bugcam detect start --model yolov8s --output results.jsonl --duration 60 --quiet

Options:

• --model <name> - Specify which model to use (default: yolov8m)
• --output <file> - Save detections to file (JSONL format)
• --duration <minutes> - Run for specified minutes (default: run indefinitely)
• --quiet - Suppress console output

Detection output format (JSONL):

{"timestamp": "2025-12-14T10:30:45", "class": "insect", "confidence": 0.92, "bbox": [100, 200, 150, 250]}

Autostart (systemd)

Manage automatic detection on system boot.

# Enable autostart
bugcam autostart enable

# Disable autostart
bugcam autostart disable

# Check status
bugcam autostart status

# View logs
bugcam autostart logs

# Follow logs in real-time
bugcam autostart logs --follow

Options:

• --follow / -f - Follow log output in real-time

Hotspot Setup

Configure the Raspberry Pi as a wireless access point for field deployments.

Create hotspot

# Create hotspot connection
sudo nmcli connection add con-name wlan-sg1 type wifi ifname wlan0 autoconnect yes ssid wlan-sg1 802-11-wireless.mode ap 802-11-wireless.band bg ipv4.method shared

# Secure with password (min 8 characters)
sudo nmcli connection modify wlan-sg1 wifi-sec.key-mgmt wpa-psk
sudo nmcli connection modify wlan-sg1 wifi-sec.psk "yourpassword"

# Activate hotspot
sudo nmcli connection up wlan-sg1

Connect to hotspot

Scan for the network SSID (e.g., wlan-sg1) and use the password to connect. Access via SSH:

ssh [e-mail]

Additional commands

# Stop hotspot
sudo nmcli connection down wlan-sg1

# Enable autostart on boot
sudo nmcli connection modify wlan-sg1 connection.autoconnect yes

Monitoring

Hailo hardware monitoring

# In terminal 1: start Hailo monitor
hailortcli monitor

# In terminal 2: enable monitoring and run detection
export HAILO_MONITOR=1
bugcam detect start

System monitoring

# CPU and memory
top  # or: btop, htop

# Temperature
vcgencmd measure_temp

# Continuous temperature monitoring
watch -n 1 vcgencmd measure_temp

# Service logs
bugcam autostart logs --follow

Performance tips

• Use yolov8s for faster inference if accuracy allows
• Use active cooling to prevent thermal throttling
• Use official Raspberry Pi power supply (5V 5A for Pi 5 + AI HAT+)

Troubleshooting

Command not found after installation

For pip install:

echo 'export PATH="$HOME/.local/bin:$PATH"' >> ~/.bashrc
source ~/.bashrc

For pipx install:

pipx ensurepath
# Close and reopen terminal

Camera not detected

# Test camera
rpicam-hello

# If fails, enable in raspi-config
sudo raspi-config
# Interface Options > Camera > Enable
sudo reboot

Hailo driver errors

# Verify Hailo installation
dpkg -l | grep hailo

# Reinstall if needed
sudo apt install --reinstall hailo-all

# Check if Hailo device is available
hailortcli scan

Service not starting

# Check service logs
bugcam autostart logs

# Verify systemd service status
systemctl status bugcam-detect.service

Models not found

# Verify models are downloaded
bugcam models list

# Download if needed
bugcam models download yolov8m

# Check cache directory
ls -lh ~/.cache/bugcam/models/

Development

For contributors who want to modify bugcam:

# Clone the repository
git clone https://github.com/MIT-Senseable-City-Lab/sensing-garden.git
cd sensing-garden

# Install with dev dependencies
poetry install

# Run tests
poetry run pytest tests/ -v

# Run CLI from source
poetry run bugcam --help

Note: Models in resources/ directory are for development only and not included in package installations.

License

MIT