Rekognition Custom Labels Max Results

Question

Hi,  
  
The AWS Rekognition Service has a limit of 30 Bounding boxs per Image When use Detect Custom Labels Action?  
I have tried to change the MaxResults parameter to other value (50) but i still getting the Maximum of 30 My custom Labeled detected on every Image that i use. And it's supposed the Rekognition Model detect more than 30 in the images that i'm using. At first i though the model needs more training to detect more my custom label.  
But this happen to every image i use, so i'm starting thinking if this is a Limit.  
I'm using confidence = 1%.  
  
Regards,  
Fábio Santos

Answer

Hi,  
  
Thanks for the answer. This 50  it's a hard limit? I want analyze images with more than 50 Labels. It's possible increase this limit or any workaround?  
  
Regards,  
Fábio Santos

Answer

Hi Fábio,  
    MaxResults does not control bounding box, please see more details in this page: https://docs.aws.amazon.com/rekognition/latest/dg/API_DetectCustomLabels.html#API_DetectCustomLabels_RequestSyntax  
  
    And the result may have up to 50 bounding boxes.

Answer

Hi Fábio,  
  
- Maximum number of unique labels per manifest is 250. That means, you can train a model that can classify/detect 250 unique labels.  
- Maximum number of labels per image is 50. That means, you can add maximum 50 labels/bounding boxes to your each image.  
  
Please refer this public documentation for more details.  
https://docs.aws.amazon.com/rekognition/latest/customlabels-dg/limits.html

Rekognition Custom Labels Max Results

관련 콘텐츠