TensorPipe: A High performance and flexible data pipeline written in core TensorFlow.

Why let your precious and intelligent model wait for the silly Data.

github.com/kartik4949/TensorPipe

Author.

Kartik Sharma, Im the creator of TensorPipe.

LinkedIn: https://www.linkedin.com/in/kartik-sharma-aaa021169/

GitHub: github.com/kartik4949

Lets Dive into TensorPipe,

Developing DataPipeline which is flexible and high performing is a painful task in TensorFlow.

State of Art Augmentations are written in a very clumsy way and it is not necessary you will find a version of it in Tensorflow.

TensorPipe fills this “Gap” with high performance, flexible and loaded with all SOTA augmentations available in the computer vision domain in one package written in core TensorFlow.

**Github Link: **github.com/kartik4949/TensorPipe

TensorPipe supports Object Detection, Keypoints, classification, and Segmentation dataset.

Some List of Augmentations:

MixUp: Mixes two randomly sampled image into one with a given alpha value.

CutMix: Completely overlaps one resized image on another random image.

**Mosaic: **Builds a mosaic-like image with randomly sampled four images.

**GridMask: **Masks grid-like masking onto the given input image.

**RandomErase: **Erases random rectangular shape from the given image.

#augmentation #keras #data-augmentation #data-pipeline #data-science