Quotas
The following quotas apply to Vertex AI Vision. These quotas apply to each Google Cloud console project and are shared across all applications and IP addresses using that project.
If you would like to process more requests, submit a Vertex AI Vision quota request for your project in the Google Cloud console.
Please provide information about your specific needs and use case in the request.
General quotas
The following general quotas apply when you use Vertex AI Vision:
| Quota | Value |
|---|---|
| API requests per minute per project | 1200 |
Stream quotas
The following quotas apply when using the Streams ingestion feature of Vertex AI Vision:
| Quota | Value |
|---|---|
| Streams per project per region | 1000 |
| Events per project per region | 20,000 |
| Analyses per project per region | 3000 |
| Series per project per region | 500,000 |
| Create event requests per minute per project per region | 600 |
| Sent packets per second per series | 100 |
| Sent bytes per second per series | 10MB |
Application platform quotas
The following quotas apply when you create and use an app using Vertex AI Vision's App platform:
| Quota | Value |
|---|---|
| Application created per project per region | 100 |
| Application deployed per project per region | 10 |
| Processor(Custom Model) per project per region | 100 |
| Instance created per project per region | 1000 |
| Instance Deployed per project per region | 100 |
Vision Warehouse quotas
The following quotas apply to Vision Warehouse. Unless explicitly specified, quotas are per project, per region.
Vision Warehouse general quotas per project:
| Quota | Value |
|---|---|
| Warehouse API requests per minute | 3,600 |
| Ingested* user specified annotation size per minute | 15 MB |
* This quota applies to annotations ingested using
CreateAnnotation and UpdateAnnotation APIs.
Streaming Warehouse specific quotas per project:
| Quota | Value |
|---|---|
| SearchAssets API requests per minute | 120 |
| Ingested media data per minute | 1,000 MB |
Warehouse storage and analysis quotas per project, per region:
| Quota | Value |
|---|---|
| Number of images | 100,000,000 |
| Number of batch videos | 5,000,000 |
| Total duration of videos | 2,160,000 minutes |
| Concurrent upload video operations | 100 |
| Concurrent analyze (video) asset operations | 100 |
| Concurrent analyze (image) corpus operations | 5 |
Warehouse index and search quotas per project per region:
| Quota | Value |
|---|---|
| Concurrent create index operations | 5 |
| Concurrent update index operations | 5 |
| Concurrent IndexAsset and RemoveIndexAsset operations | 100 |
| Search index endpint requests per minute | 6,000 |
Limitations
The following limitations apply for the Preview release of Vertex AI Vision, and are subject to change in subsequent version releases.
| Component | Limitation |
|---|---|
| Streams | Stream ingestion doesn't support a media type change (video codec, resolution, etc.) after ingestion starts. |
| Streams | The vaictl command line tool does not reconnect in the event that an ingestion stream RPC pauses. |
| Streams | Ingestion video requirements: H.264 encoding, < 1080p resolution, 25 FPS, no audio (audio is removed). |
| Streams / Warehouse | A stream can connect to only one warehouse resource. |
| Warehouse | Warehouse ingestion doesn't support media type change (video codec, resolution, etc.) after ingestion starts. |
Vision Warehouse limits
The following limits apply to Vision Warehouse. Unless explicitly specified, limits are for a given project in each region.
Warehouse storage limits:
| Limit | Value |
|---|---|
| Maximum video length | 10 hours |
| Maximum video file size | 10 GB |
| Maximum image size | 10 MB |
Number of images per ImportAssets operation |
10,000,000 |
| Maximum annotation size | string: 1 MB, custom struct: 2 MB |
| Number of asset-level annotations per asset | 1,000 |
| Number of partition-level annotations per video timestamp | 1,000 |
| Total partition-level annotation size per video timestamp | 5 MB |
| Number of exact search leaf-level* data schemas | 10,000 |
| Number of smart search leaf-level* data schemas | 100 |
* A data schema of CUSTOMIZED_STRUCT type can have multiple leaf-level data
schemas. A data schema of non-CUSTOMIZED_STRUCT type is counted as one
leaf-level data schema.
Warehouse index and search limits:
| Limit | Value |
|---|---|
| Number of index endpoints | 15 |
| Number of indexes | 100 |
| Search request text query length | 1 KB |
| Number of filter criteria in search request | 10 |
| Total length of text filter criteria | 1 KB |
| Number of search configurations | 1,000 |
| Number of search hypernyms | 1,000 |
| Number of mapped facet fields | 3 |
| Number of mapped search criteria fields | 3 |
| Number of hyponyms per hypernym | 8 |
| Number of hypernym expansions in search request* | 10 |
* One search request can contain multiple hypernyms and corresponds to more than 10 expanded requests. In this case, only the first 10 expansions are used in search.