Skip to main content
Camera parts capture images from various camera hardware. All cameras inherit from BaseCamera and provide a consistent interface for the vehicle.

ā€‹
Available Cameras

ā€‹
PiCamera (Raspberry Pi Camera Module)

The standard camera for Raspberry Pi using the Picamera2 library (Bullseye/Bookworm). Location: donkeycar/parts/camera.py:23 
from donkeycar.parts.camera import PiCamera

cam = PiCamera(
    image_w=160,      # Image width
    image_h=120,      # Image height
    image_d=3,        # Image depth (3=RGB, 1=grayscale)
    vflip=False,      # Vertical flip
    hflip=False       # Horizontal flip
)
Parameters:
  • image_w: Image width in pixels (default: 160)
  • image_h: Image height in pixels (default: 120)
  • image_d: Color depth - 3 for RGB, 1 for grayscale (default: 3)
  • vflip: Vertical flip boolean (default: False)
  • hflip: Horizontal flip boolean (default: False)
Hardware Setup:
  • Connect camera ribbon cable to CSI port on Raspberry Pi
  • Enable camera interface: sudo raspi-config ā†’ Interface Options ā†’ Camera
  • Requires: picamera2 library (pre-installed on recent Raspberry Pi OS)

ā€‹
Webcam

USB webcam support using pygame. Location: donkeycar/parts/camera.py:85 
from donkeycar.parts.camera import Webcam

cam = Webcam(
    image_w=160,
    image_h=120,
    image_d=3,
    framerate=20,
    camera_index=0    # Index if multiple cameras
)
Parameters:
  • image_w, image_h, image_d: Image dimensions
  • framerate: Target frames per second (default: 20)
  • camera_index: Camera index when multiple USB cameras (default: 0)
Installation: 
sudo apt-get install libsdl2-mixer-2.0-0 libsdl2-image-2.0-0 libsdl2-2.0-0
pip install pygame

ā€‹
CSI Camera (NVIDIA Jetson)

For Jetson Nano IMX219 camera modules using GStreamer. Location: donkeycar/parts/camera.py:194 
from donkeycar.parts.camera import CSICamera

cam = CSICamera(
    image_w=224,
    image_h=224,
    image_d=3,
    capture_width=3280,
    capture_height=2464,
    framerate=60,
    gstreamer_flip=0  # 0=none, 1=rotate CCW 90, 2=flip vert, 3=rotate CW 90
)
Parameters:
  • capture_width, capture_height: Native camera resolution
  • image_w, image_h: Output resolution (will be scaled)
  • framerate: Target FPS (default: 60)
  • gstreamer_flip: Image rotation/flip method

ā€‹
V4L Camera (Video4Linux)

Direct V4L2 access for low-level camera control. Location: donkeycar/parts/camera.py:273 
from donkeycar.parts.camera import V4LCamera

cam = V4LCamera(
    image_w=160,
    image_h=120,
    image_d=3,
    framerate=20,
    dev_fn="/dev/video0",
    fourcc='MJPG'
)
Installation: 
sudo apt-get install libv4l-dev
pip install v4l2capture

ā€‹
Advanced Cameras

ā€‹
OAK-D (Luxonis)

Intel Movidius-based depth camera with onboard AI processing. Location: donkeycar/parts/oak_d.py:34 
from donkeycar.parts.oak_d import OakD

cam = OakD(
    width=640,
    height=480,
    enable_rgb=True,
    enable_depth=True,
    device_id=None  # Or specific device serial number
)
Features:
  • RGB color camera
  • Stereo depth sensing
  • Onboard neural network inference
  • 4K video capability
Installation: 
pip install depthai
# Add udev rules:
echo 'SUBSYSTEM=="usb", ATTRS{idVendor}=="03e7", MODE="0666"' | sudo tee /etc/udev/rules.d/80-movidius.rules
sudo udevadm control --reload-rules && sudo udevadm trigger
Usage: 
V.add(cam, outputs=['cam/image_array', 'cam/depth_array'], threaded=True)

ā€‹
RealSense T265 (Intel)

Tracking camera with visual-inertial odometry. Location: donkeycar/parts/realsense2.py:13 
from donkeycar.parts.realsense2 import RS_T265

cam = RS_T265(
    image_output=False,      # Enable fisheye image output
    calib_filename=None      # Wheel odometry calibration file
)
Outputs:
  • Position (x, y, z) in meters
  • Velocity (vx, vy, vz) in m/s
  • Acceleration (ax, ay, az) in m/sĀ²
  • Optional fisheye camera images
Installation: 
pip install pyrealsense2
Usage: 
V.add(cam, 
      inputs=['enc/velocity'],
      outputs=['pos', 'vel', 'acc', 'img'], 
      threaded=True)

ā€‹
Leopard Imaging Camera

High-performance camera for NVIDIA Jetson with built-in fast stretch. Location: donkeycar/parts/leopard_imaging.py:7 
from donkeycar.parts.leopard_imaging import LICamera

cam = LICamera(
    width=224,
    height=224,
    capture_width=1280,
    capture_height=720,
    fps=60
)
Features:
  • Hardware-accelerated image processing
  • High frame rates (up to 60 FPS)
  • Optimized for Jetson platform

ā€‹
Testing Cameras

Use the camera test script: 
donkey makemovie
Or test in Python: 
from donkeycar.parts.camera import PiCamera
import time

cam = PiCamera()
time.sleep(2)  # Warm up

for i in range(10):
    img = cam.run()
    print(f"Frame {i}: shape={img.shape}")
    time.sleep(0.1)

cam.shutdown()

ā€‹
Mock Cameras for Testing

ā€‹
MockCamera

Returns a static image for testing without hardware. Location: donkeycar/parts/camera.py:333 
from donkeycar.parts.camera import MockCamera
import numpy as np

# Use with default blank image
cam = MockCamera(image_w=160, image_h=120)

# Or provide a specific image
test_image = np.random.randint(0, 255, (120, 160, 3), dtype=np.uint8)
cam = MockCamera(image=test_image)

ā€‹
ImageListCamera

Replays images from a tub for testing. Location: donkeycar/parts/camera.py:350 
from donkeycar.parts.camera import ImageListCamera

cam = ImageListCamera(path_mask='~/mycar/data/**/images/*.jpg')

ā€‹
Camera Configuration

In myconfig.py: 
# Camera type
CAMERA_TYPE = "PICAM"  # Options: PICAM, WEBCAM, CVCAM, CSIC, MOCK

# Image dimensions
IMAGE_W = 160
IMAGE_H = 120
IMAGE_DEPTH = 3  # 1 for grayscale, 3 for RGB

# Camera-specific settings
CAMERA_FRAMERATE = 20
CAMERA_VFLIP = False
CAMERA_HFLIP = False

ā€‹
Common Issues

ā€‹
Camera Not Detected

Raspberry Pi: 
# Check if camera is detected
vcgencmd get_camera
# Should show: supported=1 detected=1

# Enable camera interface
sudo raspi-config
# Navigate to Interface Options ā†’ Camera ā†’ Enable
Jetson: 
# Check for video devices
ls /dev/video*

# Test GStreamer pipeline
gst-launch-1.0 nvarguscamerasrc ! nvoverlaysink

ā€‹
Low Frame Rate

  • Reduce image resolution
  • Disable unnecessary image processing
  • Use hardware acceleration when available
  • Check USB bandwidth (for USB cameras)

ā€‹
Image Quality Issues

  • Adjust lighting conditions
  • Clean camera lens
  • Check focus (if adjustable)
  • Verify correct image orientation (use flip settings)

ā€‹
Next Steps

Build docs developers (and LLMs) love

Get started for freeTalk to us