donkey ui

​

Usage

donkey ui

​

Overview

• Visual tub management and data review
• Model training with progress visualization
• Configuration editing
• Data visualization and analysis
• Model comparison and evaluation

​

Features

​

Tub Management

• Browse and view recorded tubs
• Preview images from tub data
• Review driving sessions
• Delete or archive old tubs
• View tub metadata (record count, date, etc.)

​

Model Training

• Select training data (tubs)
• Configure training parameters
• Start training with progress bar
• View training metrics in real-time
• Save trained models

​

Data Visualization

• Plot steering and throttle distributions
• View histograms of recorded data
• Compare model predictions to user inputs
• Visualize training/validation loss curves

​

Configuration

• Edit config parameters
• Adjust model settings
• Configure hardware options
• Save configuration changes

​

Launching the UI

cd ~/mycar
donkey ui

​

Example Session

$ donkey ui
Starting Donkeycar UI...
Loading configuration from: ./config.py
UI server running at: http://localhost:8887
Open your browser to: http://localhost:8887

​

Use Cases

​

Quick Data Review

donkey ui
# Navigate to tub browser
# Click through images to review quality

​

Interactive Training

donkey ui
# Select training tab
# Choose tubs
# Configure parameters
# Click "Train"
# Watch progress in real-time

​

Configuration Editing

donkey ui
# Navigate to config editor
# Adjust parameters with sliders/inputs
# Save changes

​

Model Comparison

donkey ui
# Load model 1
# View predictions on test data
# Switch to model 2
# Compare results side-by-side

​

Benefits Over CLI

​

Visual Feedback

• See images directly instead of file paths
• Interactive plots and charts
• Real-time training progress
• Visual model comparison

​

Easier for Beginners

• No need to remember command syntax
• Guided workflows
• Form-based configuration
• Built-in help and tooltips

​

Batch Operations

• Select multiple tubs with checkboxes
• Bulk delete old data
• Compare multiple models at once

​

When to Use CLI Instead

• Automation: Scripting and batch processing
• Remote access: SSH sessions without display
• Advanced options: Full access to all parameters
• Performance: Lower overhead for large operations
• Documentation: Easier to document exact commands
• Reproducibility: Exact command history

​

Requirements

pip install donkeycar[ui]

pip install flask pillow matplotlib

​

Troubleshooting

​

UI won’t start

• Check that required dependencies are installed
• Verify no other service is using the port
• Look for error messages in terminal output
• Try running from car directory (where config.py exists)

​

Browser doesn’t open automatically

• Manually navigate to the URL shown in terminal (typically http://localhost:8887)
• Check if your firewall is blocking the port
• Try a different browser

​

Can’t see tubs or models

• Ensure you’re running from the correct directory
• Verify data/ and models/ directories exist
• Check file permissions
• Refresh the browser page

​

UI is slow or unresponsive

• Close other applications to free resources
• Try with smaller tubs first
• Check if training is running in background
• Restart the UI

​

Port already in use

Error: Port 8887 is already in use

• Stop other instances of the UI
• Kill the process using that port: lsof -ti:8887 | xargs kill
• Change the port in configuration

​

Remote access issues

• For remote access, you may need to bind to 0.0.0.0 instead of localhost
• Use SSH port forwarding: ssh -L 8887:localhost:8887 user@car-ip
• Configure firewall to allow the port

​

Comparison with Other Tools

​

Advanced Usage

​

Custom UI Port

export DONKEY_UI_PORT=9000
donkey ui

​

Multiple Cars

# Car 1
cd ~/car1
donkey ui

# Car 2 (in different terminal, different port)
cd ~/car2  
export DONKEY_UI_PORT=8888
donkey ui

​

Remote Development

# On your local machine
ssh -L 8887:localhost:8887 [email protected]

# On the remote car (in SSH session)
cd ~/mycar
donkey ui

# Now open http://localhost:8887 on your local machine

​

Best Practices

​

Security

• Don’t expose the UI to public internet without authentication
• Use SSH tunneling for remote access
• Keep Donkeycar updated for security patches

​

Performance

• Close UI when not in use to free resources
• Work with smaller datasets for faster response
• Train models via CLI for better performance

​

Workflow

• Use UI for exploration and quick tasks
• Use CLI for production workflows and automation
• Document important operations with CLI commands

​

Alternative Interfaces

​

Web Interface During Driving

python manage.py drive
# Access web interface at http://<car-ip>:8887

​

Jupyter Notebooks

import donkeycar as dk
from donkeycar.parts.tub_v2 import Tub

tub = Tub('./data/tub_1', read_only=True)
# Interactive exploration and visualization

​

Next Steps

1. Learn CLI equivalents: Understand command-line versions for automation
2. Combine approaches: Use UI for exploration, CLI for production
3. Create workflows: Document your process using both tools
4. Contribute: Report UI bugs or suggest features to Donkeycar project

• donkey createcar - Create new car projects
• donkey train - Train models with full options
• donkey tubplot - Advanced prediction analysis
• donkey makemovie - Create detailed videos