Image Segmentation

Image Segmentation divides an image into segments where each pixel in the image is mapped to an object.

For more details about the image-segmentation task, check out its dedicated page! You will find examples and related materials.

Recommended models

openmmlab/upernet-convnext-small: Solid semantic segmentation model trained on ADE20k.
facebook/mask2former-swin-large-coco-panoptic: Panoptic segmentation model trained on the COCO (common objects) dataset.

Explore all available models and find the one that suits you best here.

Using the API

Language

Client

Provider

Settings

from huggingface_hub import InferenceClient

client = InferenceClient(
    provider="hf-inference",
    api_key="hf_xxxxxxxxxxxxxxxxxxxxxxxx",
)

output = client.image_segmentation("cats.jpg", model="jonathandinu/face-parsing")

API specification

Request

Headers
authorization	string	Authentication header in the form `'Bearer: hf_**'` when `hf_**` is a personal user access token with “Inference Providers” permission. You can generate one from your settings page.

Payload
inputs*	string	The input image data as a base64-encoded string. If no `parameters` are provided, you can also provide the image data as a raw bytes payload.
parameters	object
mask_threshold	number	Threshold to use when turning the predicted masks into binary values.
overlap_mask_area_threshold	number	Mask overlap threshold to eliminate small, disconnected segments.
subtask	enum	Possible values: instance, panoptic, semantic.
threshold	number	Probability threshold to filter out predicted masks.

Response

Body
(array)	object[]	A predicted mask / segment
label	string	The label of the predicted segment.
mask	string	The corresponding mask as a black-and-white image (base64-encoded).
score	number	The score or confidence degree the model has.