This topic describes the 2023 product and documentation updates for the Alibaba Cloud Visual Intelligence API.
May 2023
Category | Capability Name | Description | Release date | Region | References |
Image analysis and processing | Gastric cancer detection | Detects gastric cancer and non-gastric cancer lesions from input CT scans that cover the stomach, such as chest or abdominal CT scans. | 2023-05-26 | China (Shanghai) |
April 2023
Category | Capability Name | Description | Release date | Region | References |
Image analysis and processing | Fatty liver detection | Locates and segments the liver and spleen from input chest or abdominal CT images. Then measures global or local density of the liver and spleen. Finally, determines whether fatty liver is present and its severity based on the measurements and a deep model. | 2023-04-21 | China (Shanghai) |
March 2023
Category | Capability | Description | Release date | Region | References |
Asynchronous task management | Query asynchronous task results | For asynchronous API operations, the initial response does not contain the actual result. Save the RequestId from the response. Then call GetAsyncJobResult to retrieve the actual result. | 2023-03-02 | China (Shanghai) | |
Query asynchronous task list | After you submit an asynchronous task, call QueryAsyncJobList to retrieve a list of submitted tasks. | 2023-03-02 | China (Shanghai) | ||
Cancel queued asynchronous tasks | Call CancelWaitingAsyncJob to cancel asynchronous tasks with status QUEUING. You cannot cancel tasks with status PROCESSING. After cancellation, the task status changes to JOB_CANCELED. Canceled tasks will not run again. | 2023-03-02 | China (Shanghai) | ||
Image analysis and processing | Radiation therapy lymph node segmentation | Segments lymph nodes from input chest CT scans, either non-contrast or contrast-enhanced. Specify the target area as CHEST. Returns results as NIFTI format masks. | 2023-03-21 | China (Shanghai) | |
Bone mineral density estimation | Locates and labels vertebrae and estimates bone mineral density from input chest or abdominal CT images. | 2023-03-31 | China (Shanghai) |
February 2023
Category | Capability Name | Description | Release date | Region | References |
Image analysis and processing | Radiation therapy target delineation | Automatically delineates radiation therapy targets from input chest CT scans, either non-contrast or contrast-enhanced. Specify the cancer type and target type. | 2023-02-01 | China (Shanghai) | |
Face and body | Infrared face liveness detection | Detects whether a face in an infrared image is a live, bare-faced person captured up close by an authenticated device. This capability supports real-time infrared face capture scenarios and meets security and authenticity requirements for infrared face registration and authentication. | 2023-02-02 | China (Shanghai) | |
Masked face comparison (1:1) | Compares the largest face detected in each of two input images to determine whether they belong to the same person. Uses three optimized technologies—mask generation, occlusion-resistant keypoint localization, and occlusion-resistant feature attention—to enable fast face recognition while wearing a mask. | 2023-02-02 | China (Shanghai) | ||
Image generation | Generative image cartoonization | Converts an input image into a cartoon-style image at the same resolution. Select the desired cartoon style before generating. | 2023-02-08 | China (Shanghai) | |
Generative image super resolution | Enhances image details, repairs damaged areas, and upscales images. Improves detail richness and clarity. | 2023-02-17 | China (Shanghai) |
January 2023
Category | Capability Name | Description | Release date | Region | References |
Image generation | Text-to-image | Uses a DAMO Academy-developed text-to-image Large Language Model (LLM). Applies knowledge recombination and a variable-dimension diffusion model to accelerate convergence and improve output quality. Input descriptive text to generate a 2D image matching the description. Supports both Chinese and English input. | 2023-01-11 | China (Shanghai) | Unpublished |
Image generation | Text-and-image-to-image | Uses a DAMO Academy-developed text-to-image Large Language Model (LLM). Input descriptive text and a reference image to generate a new image that matches both the text and the visual features of the reference image. This improves control over the output. | 2023-01-11 | China (Shanghai) | Unpublished |