Introduction

Capybara consists of the following parts (based on the current code structure):

• Vision (capybara.vision): Image/video I/O and processing.
• Structures (capybara.structures): Geometric structures such as Box/Boxes, Polygon/Polygons, and Keypoints.
• Runtime (capybara.runtime): Registry and selection logic for runtime/backend.
• Inference engines (optional):
  • capybara.onnxengine (ONNXRuntime)
  • capybara.openvinoengine (OpenVINO)
  • capybara.torchengine (TorchScript)
• Utils (capybara.utils): Utilities (paths, downloads, time, etc.).
• Extras (optional):
  • visualization: drawing utilities (capybara.vision.visualization)
  • ipcam: simple Web demo (capybara.vision.ipcam)
  • system: system info utilities (capybara.utils.system_info)

Vision

The Vision module focuses on image/video processing and I/O.

Directory structure:

vision
├── functionals.py       # Basic image processing functions, such as filtering and transformations
├── geometric.py         # Geometric processing functions, such as rotation and scaling
├── improc.py            # Core image processing logic
├── morphology.py        # Morphological processing functions, such as dilation and erosion
├── videotools           # Video-related tools
├── ipcam                # (extra: ipcam) IPCam demo
└── visualization        # (extra: visualization) drawing/visualization

Main features:

• Image/video reading, processing, and visualization.
• Supports multiple sources (local files, video frame extraction, IPCam demo, etc.).

Structures

The Structures module is responsible for handling structured data, commonly used in computer vision and data analysis scenarios.

Directory structure:

structures
├── functionals.py       # Related functional functions
├── boxes.py             # Box and Boxes data structures
├── keypoints.py         # Keypoints data structure
└── polygons.py          # Polygon and Polygons data structures

Main features:

• Provides structured data processing for Boxes, Keypoints, and Polygons.
• Supports operations such as intersection, IoU, scaling, etc.

Runtime / Inference engines (optional)

Inference-related features live in separate modules, and capybara.runtime provides a unified way to describe/choose runtime/backend.

Note: inference backends are optional dependencies. Install the required extras first (e.g. capybara-docsaid[onnxruntime]).

capybara.runtime

• Defines Runtime / Backend, and provides selection helpers such as auto_backend_name().

capybara.onnxengine

Directory structure:

onnxengine
├── engine.py            # Core inference logic
├── __init__.py          # Initialization file
├── metadata.py          # Model metadata management
└── utils.py             # ONNX helpers

Main features:

• Load and run inference on ONNX models.

capybara.openvinoengine

• OpenVINO inference wrapper (sync inference + optional async queue).

capybara.torchengine

• TorchScript inference wrapper (simple dtype/device normalization).

Utils

The Utils module contains a wide range of auxiliary utility functions.

Directory structure:

utils
├── custom_path.py       # Custom path operations
├── custom_tqdm.py       # Progress bar utility
├── files_utils.py       # File handling functions
├── powerdict.py         # Enhanced dictionary operations
├── system_info.py       # System information detection
├── time.py              # Time handling utilities
└── utils.py             # General utility functions

Main features:

• File handling and system information detection.
• Provides custom tools like progress bars and enhanced dictionaries.

Tests

The Tests module is used to verify that the system functions correctly.

Main features:

• Includes unit tests for various modules.
• Provides fast regression and functionality verification.

---

This is a high-level introduction to the modules of Capybara.

For API usage and examples, continue reading the docs. If you see an import error, check whether the corresponding extra is installed.