ETA (Extensible Toolkit for Analytics)

Common errors

• ModuleNotFoundError: No module named 'eta'

  cause The `voxel51-eta` package has not been installed in your current Python environment.
  fix Install the library using pip: `pip install voxel51-eta`

• TypeError: fromstring() was called with a non-string type.

  cause You are likely using an older version of `voxel51-eta` (prior to `0.15.3`) with `numpy>=2.0`, which deprecated `numpy.fromstring` in favor of `numpy.frombuffer` for bytes-like objects.
  fix Upgrade `voxel51-eta` to `0.15.3` or newer to fix `numpy` compatibility issues: `pip install --upgrade voxel51-eta`

• AttributeError: module 'eta.core.datasets' has no attribute 'ETADataset'

  cause You are attempting to access a specific class or function (`ETADataset` in this case) that might have been renamed, moved, or removed in your version of `eta`. As an internal toolkit, its API can be less stable and more subject to change.
  fix Consult the `voxel51-eta` source code or `FiftyOne` documentation for the correct API. For typical use cases, consider using `FiftyOne` directly as it provides a stable and documented interface.

Warnings

• gotcha The `voxel51-eta` library is primarily an internal toolkit developed for Voxel51's FiftyOne computer vision framework. While public on PyPI, it is not typically used as a standalone library by end-users. Its APIs might be less stable or documented for independent development compared to external-facing libraries.
  Fix: For general computer vision tasks and analytics, consider using the higher-level `fiftyone` library directly, which abstracts away much of `eta`'s underlying complexity.
• breaking Python 2 compatibility was completely removed in `v0.15.0`. The library now strictly requires Python 3.9 or newer.
  Fix: Ensure your Python environment is running Python 3.9 or a later version before installing or upgrading `voxel51-eta` to version `0.15.0` or higher.
• gotcha Compatibility issues with `numpy>=2.0` were addressed in `v0.15.3`, specifically regarding video processing operations that used `numpy.fromstring` (now `numpy.frombuffer`). If you encounter errors related to `fromstring()` with recent NumPy versions, your `eta` version might be too old.
  Fix: Update `voxel51-eta` to version `0.15.3` or higher by running `pip install --upgrade voxel51-eta` to ensure compatibility with `numpy>=2.0`.

Install

• pip install voxel51-eta Install stable version

Imports

• Config

  from eta.core.config import Config

  For loading and managing ETA configurations.
• ETADataset

  from eta.core.datasets import ETADataset

  Represents a generic ETA dataset structure.
• get_eta_data_dir

  from eta.core.utils import get_eta_data_dir

  Utility to get the base data directory for ETA.

Quickstart

This quickstart shows how to access the default data directory used by ETA, demonstrating a basic utility function. As an internal toolkit, extensive standalone examples are less common, but core utilities are accessible.

import os
from eta.core.utils import get_eta_data_dir

# ETA is primarily an internal toolkit for FiftyOne. 
# This quickstart demonstrates a basic utility function.

try:
    eta_data_dir = get_eta_data_dir()
    print(f"ETA data directory: {eta_data_dir}")
    
    if not os.path.exists(eta_data_dir):
        print(f"Creating ETA data directory at: {eta_data_dir}")
        os.makedirs(eta_data_dir)
    
    # Example: creating a dummy file in the ETA data directory
    dummy_file_path = os.path.join(eta_data_dir, "quickstart_dummy.txt")
    with open(dummy_file_path, "w") as f:
        f.write("This is a dummy file created by ETA quickstart.")
    print(f"Dummy file created at: {dummy_file_path}")
    
except Exception as e:
    print(f"An error occurred: {e}")
    print("ETA is an internal toolkit; direct usage might require deeper knowledge of its integration with FiftyOne.")