https://github.com/beringresearch/datasets

Package for flexible tensorflow dataset creation
https://github.com/beringresearch/datasets

dataset-generation deeplearning tensorflow

Last synced: 2 months ago
JSON representation

Package for flexible tensorflow dataset creation

• Host: GitHub
• URL: https://github.com/beringresearch/datasets
• Owner: beringresearch
• License: apache-2.0
• Created: 2020-08-05T08:34:59.000Z (almost 6 years ago)
• Default Branch: master
• Last Pushed: 2022-03-18T11:33:56.000Z (over 4 years ago)
• Last Synced: 2025-09-09T15:06:11.415Z (10 months ago)
• Topics: dataset-generation, deeplearning, tensorflow
• Language: Python
• Homepage:
• Size: 18.7 MB
• Stars: 1
• Watchers: 4
• Forks: 0
• Open Issues: 5
• Metadata Files:
  • Readme: README.md
  • License: LICENSE

Awesome Lists containing this project

README

          
# Build Tensorflow Input Pipelines

`datasets` is a python package that enables users to quickly build complex Tensorflow datasets. The tool offers flexibility to import out-of-memory datasets and apply image augmentation functions in real time.

`datasets` API borrows heavily from [`ImageDataGenerator`](https://www.tensorflow.org/api_docs/python/tf/keras/preprocessing/image/ImageDataGenerator), making it nearly a drop-in replacement. However, `TFImageDataset` class is approximately 5-fold faster than the `ImageDataGenerator`.

## Installation

The latest stable version can be installed directly from github:

```bash

git clone https://github.com/beringresearch/datasets/

cd datasets

python3 install --editable .

```

## Getting Started

Check out [example notebook](https://github.com/beringresearch/datasets/blob/master/examples/Import%20Images%20from%20DataFrame.ipynb) to get started with the package.