Template: Image Embedding Pipeline

Authors:

Overview

Note: this is just a template, not a runnable pipeline.

This pipeline cannot be run, which is the template for the image embedding pipeline class and defines YAML template file for embedding images, as well as the standard inputs and outputs. You can complete the pipeline by filling in the parameters(init_args) of the Operator section in image_embedding_pipeline_template.yaml and update this README file. FYI, image-embedding-resnet50 is based on this template.

This pipeline is used to extract the feature vector of the image. It first normalizes the image and then uses a model to generate the vector.

Interface

Input Arguments:

• img_tensor:
  • the input image to be encoded
  • supported types: PIL.Image

Pipeline Output:

The pipeline returns a tuple Tuple[('cnn', numpy.ndarray)] containing following fields:

• feature_vector:
  • the embedding of input image
  • data type: numpy.ndarray

How to use

1. Install Towhee

$ pip3 install towhee

You can refer to Getting Started with Towhee for more details. If you have any questions, you can submit an issue to the towhee repository.

2. Run it with Towhee

>>> from towhee import pipeline
>>> from PIL import Image

>>> img = Image.open('path/to/your/image') #for example './test.jpg'
>>> embedding_pipeline = pipeline('user/repo_name')  #the pipeline repo, such as 'towhee/image-embedding-resnet50'
>>> embedding = embedding_pipeline(img)

How it works

This pipeline includes two main operators: transform image and image embedding. The transform image operator will first convert the original image into a normalized format, such as with 512x512 resolutions. Then, the normalized image will be encoded via image embedding operator, and finally we get a feature vector of the given image.

Refer Towhee architecture for basic concepts in Towhee: pipeline, operator, dataframe.