# Pipeline: Image Embedding using Resnet50 Authors: derekdqc ## Overview The pipeline is used to **extract the feature vector of a given image**. It first normalizes the image and then uses Resnet50 model to generate the vector. ## Interface **Input Arguments:** - img_path: - the input image to be encoded - supported types: `str` (path of the image) **Pipeline Output:** The pipeline returns a tuple `Tuple[('feature_vector', numpy.ndarray)]` containing following fields: - feature_vector: - the embedding of input image - data type: `numpy.ndarray` - shape: (2048,) ## How to use 1. Install [Towhee](https://github.com/towhee-io/towhee) ```bash $ pip3 install towhee ``` > You can refer to [Getting Started with Towhee](https://towhee.io/) for more details. If you have any questions, you can [submit an issue to the towhee repository](https://github.com/towhee-io/towhee/issues). 2. Run it with Towhee ```python >>> from towhee import pipeline >>> embedding_pipeline = pipeline('towhee/image-embedding-resnet50') >>> embedding = embedding_pipeline('path/to/your/image') ``` ## How it works This pipeline includes two main operators: [transform image](https://hub.towhee.io/towhee/transform-image-operator-template) (implemented as [towhee/transform-image](https://hub.towhee.io/towhee/transform-image)) and [image embedding](https://hub.towhee.io/towhee/image-embedding-operator-template) (implemented as [towhee/resnet-image-embedding](https://hub.towhee.io/towhee/resnet-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](https://github.com/towhee-io/towhee#towhee-architecture) for basic concepts in Towhee: pipeline, operator, dataframe. ![img](./readme_res/pipeline.png)