tomo-detect 0.1.4

A CLI tool for detecting motor coordinates in tomography data

Tomo-Detect

A command-line tool for detecting motor coordinates in tomography data using deep learning models.

Installation

pip install tomo-detect

Usage

Basic usage:

tomo-detect input_path

The input path can be either:

• A .npy file containing tomography data
• A zip file containing multiple tomography files
• A directory containing .npy or .mrc files

Options

tomo-detect --help                     # Show help message and usage information
tomo-detect input.npy --debug          # Enable debug logging
tomo-detect input.zip --test           # Run test mode with additional validations
tomo-detect input.npy --output path    # Specify custom output directory
tomo-detect input.npy --batch-size 4   # Set custom batch size for inference
tomo-detect input.npy --device cpu     # Force CPU inference

Output Files

The tool generates several output files:

1. motor_detections_submission.csv - Contains motor coordinates (primary output)
2. motor_detections_detailed.csv - Includes additional detection information
3. predictions.npy - Raw probability maps
4. masks.npy - Binary masks derived from predictions
5. summary.json - Detection summary and statistics

Example

# Process a single .npy file
tomo-detect sample.npy

# Process multiple files in a zip
tomo-detect samples.zip

# Enable debug mode for detailed logging
tomo-detect sample.npy --debug

Tips and Tricks

1. Input Preparation:

   • Ensure input files are properly formatted numpy arrays
   • For .mrc files, they will be automatically converted
   • Zip files should contain only supported file types

2. Performance Optimization:

   • Use GPU acceleration when available
   • Adjust batch size based on your memory capacity
   • For large datasets, consider processing in chunks

3. Troubleshooting:

   • Enable --debug mode for detailed logging
   • Check system requirements before running
   • Verify input file formats and dimensions

System Requirements

• Python 3.8 or higher
• CUDA-capable GPU (optional but recommended)
• Minimum 8GB RAM
• 2GB disk space for models

License

MIT License