Skip to main content
The lerobot-find-cameras command detects OpenCV and Intel RealSense cameras, displays their capabilities, and captures test images.

​
Command

lerobot-find-cameras [CAMERA_TYPE] [OPTIONS]
Location: src/lerobot/scripts/lerobot_find_cameras.py

​
Overview

This utility helps you:
  • Detect all connected cameras (OpenCV and RealSense)
  • View camera capabilities (resolution, FPS, format)
  • Identify camera indices and serial numbers
  • Capture test images from all cameras
  • Verify camera functionality before recording

​
Key Options

​
camera_type
str
Optional filter: opencv or realsense. If omitted, shows all cameras.
​
--output-dir
str
default:"outputs/captured_images"
Directory to save test images.
​
--record-time-s
float
default:"6.0"
Duration in seconds to capture test images.

​
Usage Examples

​
Detect All Cameras

lerobot-find-cameras
This displays all OpenCV and RealSense cameras with their metadata:
  • Camera type and index/serial number
  • Default resolution and FPS
  • Video format (MJPG, YUYV, etc.)
  • USB connection type

​
Detect Only OpenCV Cameras

lerobot-find-cameras opencv
Example output: 
--- Detected Cameras ---
Camera #0:
  Type: OpenCV
  Id: 0
  Backend: V4L2
  Default stream profile:
    Width: 640
    Height: 480
    Fps: 30
    Format: MJPG
--------------------
Camera #1:
  Type: OpenCV
  Id: /dev/video4
  Backend: V4L2
  Default stream profile:
    Width: 1280
    Height: 720
    Fps: 30
    Format: YUYV
--------------------

​
Detect Only RealSense Cameras

lerobot-find-cameras realsense
Example output: 
--- Detected Cameras ---
Camera #0:
  Type: RealSense
  Id: 123456789
  Name: Intel RealSense D405
  Firmware version: 5.15.0.2
  Usb type descriptor: USB 3.2
  Default stream profile:
    Width: 640
    Height: 480
    Fps: 30
    Format: RGB8
--------------------

​
Capture Test Images

lerobot-find-cameras --output-dir=./camera_test --record-time-s=10
This:
  1. Detects all cameras
  2. Connects to each camera
  3. Captures images for 10 seconds
  4. Saves images to ./camera_test/
Output files are named: {camera_type}_{camera_id}.png Example: 
outputs/captured_images/
├── opencv_0.png
├── opencv_2.png
└── realsense_123456789.png

​
Filter and Capture

lerobot-find-cameras opencv --output-dir=./opencv_test --record-time-s=5
Captures images only from OpenCV cameras for 5 seconds.

​
Camera Information

​
OpenCV Cameras

For each OpenCV camera, displays:
  • Type: Always “OpenCV”
  • Id: Camera index (0, 1, 2…) or device path (/dev/video4)
  • Backend: Video capture backend (V4L2, MSMF, etc.)
  • Default stream profile:
    • Width and height (pixels)
    • FPS (frames per second)
    • Format (MJPG, YUYV, H264, etc.)

​
RealSense Cameras

For each RealSense camera, displays:
  • Type: Always “RealSense”
  • Id: Camera serial number
  • Name: Camera model (e.g., “Intel RealSense D405”)
  • Firmware version: Current firmware version
  • USB type descriptor: USB connection type
  • Default stream profile:
    • Color stream resolution and FPS
    • Depth stream resolution and FPS (if available)

​
Common Use Cases

​
Identify Camera Indices for Recording

Before recording a dataset: 
lerobot-find-cameras
Use the displayed indices in your recording configuration: 
lerobot-record \
  --robot.cameras='{
    laptop: {type: opencv, index_or_path: 0, width: 640, height: 480, fps: 30},
    phone: {type: opencv, index_or_path: 2, width: 640, height: 480, fps: 30}
  }'

​
Find RealSense Serial Numbers

lerobot-find-cameras realsense
Use the serial number in your configuration: 
lerobot-record \
  --robot.cameras='{
    realsense: {
      type: realsense,
      serial_number_or_name: 123456789,
      width: 640,
      height: 480,
      fps: 30
    }
  }'

​
Verify Camera Functionality

Test cameras before a long recording session: 
lerobot-find-cameras --output-dir=./test --record-time-s=30
Check the saved images in ./test/ to verify:
  • All cameras are detected
  • Image quality is acceptable
  • No USB bandwidth issues
  • Correct camera orientation

​
Debug Camera Issues

If cameras aren’t detected:
  1. Run without filter to see all devices: 
    lerobot-find-cameras
  2. Check if the camera appears but can’t connect: 
    lerobot-find-cameras --record-time-s=2
    Look for connection errors in the output.
  3. Try specific camera type: 
    lerobot-find-cameras opencv
lerobot-find-cameras realsense

​
Troubleshooting

​
No Cameras Detected

OpenCV cameras:
  • Check USB connections
  • Verify camera permissions (Linux: add user to video group)
  • Try different USB ports
  • Close other applications using cameras
RealSense cameras:
  • Ensure pyrealsense2 is installed: pip install pyrealsense2
  • On Linux, install udev rules: sudo apt install librealsense2-dkms
  • Use USB 3.0 ports for best performance
  • Update firmware using Intel RealSense Viewer

​
Camera Found But Can’t Capture Images

  • Camera may be in use by another application
  • Insufficient USB bandwidth (try disconnecting other USB devices)
  • Driver issues (update camera drivers)
  • Wrong USB port (use USB 3.0 for high-resolution cameras)

​
Wrong Camera Index

On Linux, camera indices can change. Use device paths instead: 
# Find actual device path
ls -l /dev/video*

# Use path instead of index
lerobot-find-cameras  # Shows paths like /dev/video4
Or create udev rules for stable paths (see Camera Support).

​
Permission Denied (Linux)

# Add user to video group
sudo usermod -a -G video $USER

# Log out and back in for changes to take effect

​
Programmatic Usage

from lerobot.scripts.lerobot_find_cameras import find_and_print_cameras
from pathlib import Path

# Find all cameras
all_cameras = find_and_print_cameras(camera_type_filter=None)

# Find only OpenCV cameras
opencv_cameras = find_and_print_cameras(camera_type_filter="opencv")

# Find only RealSense cameras
realsense_cameras = find_and_print_cameras(camera_type_filter="realsense")

# Access camera metadata
for camera in all_cameras:
    print(f"Type: {camera['type']}")
    print(f"ID: {camera['id']}")
    if 'default_stream_profile' in camera:
        profile = camera['default_stream_profile']
        print(f"Resolution: {profile['width']}x{profile['height']} @ {profile['fps']} FPS")

​
See Also

Build docs developers (and LLMs) love

Get started for freeTalk to us