Image Embedding Pipeline

Description

An image embedding pipeline generates a vector given an image. This Pipeline extracts features for image with 'ResNet50' models provided by Timm. Timm is a deep-learning library developed by Ross Wightman, who maintains SOTA deep-learning models and tools in computer vision.

Code Example

Create pipeline with the default configuration

from towhee import AutoPipes

p = AutoPipes.pipeline('image-embedding')
res = p('https://github.com/towhee-io/towhee/raw/main/towhee_logo.png')
res.get()

Create pipeline and set the configuration

More parameters refer to the Configuration.

from towhee import AutoPipes, AutoConfig

conf = AutoConfig.load_config('image-embedding')
conf.model_name = 'resnet34'

p = AutoPipes.pipeline('image-embedding', conf)
res = p('https://github.com/towhee-io/towhee/raw/main/towhee_logo.png')
res.get()

Configuration

ImageEmbeddingConfig

You can find some parameters in image_decode.cv2 and image_embedding.timm operators.

mode: str

The mode for image, 'BGR' or 'RGB', defaults to 'BGR'.

model_name: str

The model name in string. The default value is "resnet50". Refer to Timm Docs to get a full list of supported models.

num_classes: int

The number of classes. The default value is 1000. It is related to model and dataset.

skip_preprocess: bool

The flag to control whether to skip image pre-process. The default value is False. If set to True, it will skip image preprocessing steps (transforms). In this case, input image data must be prepared in advance in order to properly fit the model.

device: int

The number of GPU device, defaults to -1, which means using CPU.

Interface

Encode the image and generate embedding vectors.

Parameters:

img: str

Path or url of the image to be loaded.

Returns: np.ndarray

Embedding vectors to represent the image.