Skip to content
Model infoModelsSnapshot: 15 Oct 2026

Alibaba Cloud / Alibaba image utilities

Human Instance Segmentation

Produces a pixel-level mask that distinguishes people in an image, plus a visualization. Intended for person selection, cutouts and downstream image editing; access is a finite free trial.

restrictedClosed weightsImage segmentation
Official documentation (opens in a new tab)Reviewed 14 Oct 2026
28 of 37 specification fields collected
Image input limitInput: 512–4,096 pixels per side inclusive; at most 10 MB.See specification notes below
Output typesImageSee specification notes below
Input billingSee billing notesAlibaba Model Studio · Beijing · trial only
Output billingSee billing notesAccess and billing notes below

Specifications

Provider-documented facts. “Not collected” means we have not verified that field yet, not that the feature is unsupported.

Context / input limit
Not collected
Maximum input tokens
Not collected
Maximum output
Not collected
Reasoning
Not collected
Streaming
Not collected
Knowledge cutoff
Not collected
Parameters
Not collected
Architecture
Not collected
Languages
Not collected
API aliases
image-instance-segmentationSource for API aliases (opens in a new tab)
Published availability regions
China (Beijing)Source for Published availability regions (opens in a new tab)
Request mode
Async HTTP task submission with X-DashScope-Async: enable; query GET /api/v1/tasks/{task_id}.Source for Request mode (opens in a new tab)
Published price
Trial only; paid usage unavailable.Source for Published price (opens in a new tab)
Usage metering
500 images in the published trial; no further calls after exhaustion. Not an unrestricted free service.Source for Usage metering (opens in a new tab)
Free usage allowance
500 images per model, valid for 90 days from the latest activation, model release or application approval. Account eligibility and remaining entitlement unverified.Source for Free usage allowance (opens in a new tab)
Image input limits
Input: 512–4,096 pixels per side inclusive; at most 10 MB.Source for Image input limits (opens in a new tab)
Input image formats
JPEG, PNG, JPG, BMP, WEBPSource for Input image formats (opens in a new tab)
Output link expiry
Completed task and results are deleted after 24 hours. Output image URLs expire after 24 hours. This does not establish input retention or training-use policy.Source for Output link expiry (opens in a new tab)
Required request fields
One input.image_url; no text prompt or person-ID selection parameter is documented.Source for Required request fields (opens in a new tab)
Image URL requirements
Public HTTP(S) URL without Chinese characters. Base64 support is not documented in this contract.Source for Image URL requirements (opens in a new tab)
Generated assets
output_image_url: instance-label mask, output_vis_image_url: visualization of segmentation, usage.image_count: generated-image countSource for Generated assets (opens in a new tab)
Output resolutions
Output-mask and visualization aspect ratio: 1:10 to 10:1; exact pixel dimensions unverifiedSource for Output resolutions (opens in a new tab)
Mask behavior
Official appendix treats mask channel 0 as integer person labels starting at 1, with 0 unselected/background. Its client-side example selects reserve_ids and creates 0/255 preserve and remove masks. Those IDs are not an API request parameter.Source for Mask behavior (opens in a new tab)
Related image workflow
Generate person labels, select the desired instance IDs locally, then send binary erase/preserve masks to image-erase-completion.Source for Related image workflow (opens in a new tab)

The API appendix provides the mask extraction example.

Access & pricing

Prices belong to an endpoint and billing tier. A provider listing does not guarantee access for every account or region.

Alibaba Model Studio · Beijing · trial only

Provider source (opens in a new tab)
API model ID
image-instance-segmentation
Access
POST https://{WorkspaceId}.cn-beijing.maas.aliyuncs.com/api/v1/services/aigc/image2image/image-synthesis. Async HTTP task submission with X-DashScope-Async: enable; query GET /api/v1/tasks/{task_id}.
Regions
China (Beijing)
Published request quota
120 requests per minuteSource for Published request quota (opens in a new tab)
Task submission rate
2 submissions per secondSource for Task submission rate (opens in a new tab)
Quota scope
Per-model limit shared by the Alibaba Cloud account and RAM usersSource for Quota scope (opens in a new tab)
Request mode
Async HTTP task submission with X-DashScope-Async: enable; query GET /api/v1/tasks/{task_id}.Source for Request mode (opens in a new tab)
Free usage allowance
500 images per model, valid for 90 days from the latest activation, model release or application approval. Account eligibility and remaining entitlement unverified.Source for Free usage allowance (opens in a new tab)

Endpoint pricing has not been collected. Check the provider for applicable rates.

Data & enterprise terms

Task/result expiry documented; input retention, training use and account entitlement unverified.

Caching

Caching can reuse earlier input processing. Availability, lifetime and billing depend on the endpoint. A cache hit still generates a new response.

01 / INITIAL INPUTProcess & retain

The system processes eligible input and retains reusable state.

02 / LATER REQUESTCheck available state

Reuse depends on matching input, session state and the cache policy.

03 / CACHE HITReuse & generate

Reusing state can reduce repeated work. Any charges follow the billing rules below.

What the measurements say

No benchmark observations have been added for this model. We do not turn a vendor description into a performance score.

Latency and throughput depend on the host, request size, reasoning settings and load. There is no measured speed or global usage figure in this record.

The record behind the model

Why it is included

Qualified classification inference: separately named native Alibaba hosted model documented by its own provider card and API, with no exact downloadable checkpoint identified in those sources. Public segmentation and inpainting research is not automatically equivalent to this deployment. Exact private/public checkpoint lineage remains under review.

Classification evidence (opens in a new tab)

Lifecycle dates

Model Studio release19 Aug 2024Source (opens in a new tab)

Sources reviewed