Makes a prediction using a custom model based on the input data.
The following table resumes the configuration properties (access credentials) you must set in order to use this AI task.
|
PropertyKey |
ProviderType |
Key |
Alibaba |
- |
Amazon |
- |
Baidu |
- |
Google |
Service Account JSON |
IBM |
Visual Recognition Key |
Microsoft |
Custom Vision Training Key |
SAP |
- |
Tencent |
- |
After deploying your model (trained with the Mamaevs' Flowers Recognition dataset) and taking the following (unseen) image input, the table below shows the prediction made for each provider and the time it takes for processing it.
Note: In the context of custom models, the term "unseen image" means the image will not participate in the training process. In other words, the image is not in the training dataset.
Provider |
Output |
Benchmark |
Alibaba |
GXAI6001 - Task 'GeneXusAI.Custom.Evaluate' is unavailable (...) |
N/A |
Amazon |
GXAI6001 - Task 'GeneXusAI.Custom.Evaluate' is unavailable (...) |
N/A |
Baidu |
GXAI6001 - Task 'GeneXusAI.Custom.Evaluate' is unavailable (...) |
N/A |
Google |
{
"Label": "dandelion",
"Confidence": 0.915
}
|
3377ms |
IBM |
{
"Label": "dandelion",
"Confidence": 0.945
}
|
1070ms |
Microsoft |
{
"Label": "dandelion",
"Confidence": 0.974
}
|
590ms |
SAP |
GXAI6001 - Task 'GeneXusAI.Custom.Evaluate' is unavailable (...) |
N/A |
Tencent |
GXAI6001 - Task 'GeneXusAI.Custom.Evaluate' is unavailable (...) |
|
-
If you want to get a set of possible candidates, use a predefined GeneXusAI task instead of the Predict procedure (e.g. use Classify procedure if your model was trained for predicting Labels, or DetectObjects for predicting Regions).
This procedure is available as of GeneXus 16 upgrade 6.