You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Copy file name to clipboardExpand all lines: articles/ai-services/computer-vision/includes/how-to-guides/analyze-image-40-java.md
+3-3Lines changed: 3 additions & 3 deletions
Display the source diff
Display the rich diff
Original file line number
Diff line number
Diff line change
@@ -86,7 +86,7 @@ Create a new [ImageAnalysisOptions](/java/api/azure.ai.vision.imageanalysis.imag
86
86
87
87
### Set model name when using a custom model
88
88
89
-
You can also do image analysis with a custom trained model. To create and train a model, see [Create a custom Image Analysis model](/azure/ai-services/computer-vision/how-to/model-customization). Once your model is trained, all you need is the model's name. You don't need to specify visual features if you use a custom model.
89
+
You can also do image analysis with a custom trained model. To create and train a model, see [Create a custom Image Analysis model](/azure/ai-services/computer-vision/how-to/model-customization). Once your model is trained, all you need is the model's name.
90
90
91
91
To use a custom model, create the [ImageAnalysisOptions](/java/api/azure.ai.vision.imageanalysis.imageanalysisoptions) object and call the [setModelName](/java/api/azure.ai.vision.imageanalysis.imageanalysisoptions.modelname#azure-ai-vision-imageanalysis-imageanalysisoptions-modelname) method. You don't need to set any other properties on **ImageAnalysisOptions**. There's no need to call [setFeatures](/java/api/azure.ai.vision.imageanalysis.imageanalysisoptions.features#azure-ai-vision-imageanalysis-imageanalysisoptions-features), as you do with the standard model, since your custom model already implies the visual features the service extracts.
92
92
@@ -137,7 +137,7 @@ This section shows you how to make an analysis call to the service using the sta
137
137
138
138
1. Call the **getReason** method on the [ImageAnalysisResult](/java/api/azure.ai.vision.imageanalysis.imageanalysisresult) object, to determine if analysis succeeded or failed.
139
139
140
-
1. If succeeded, proceed to access the call relevant result methods based on your selected visual features, as shown here. Additional information (not commonly needed) can be obtained by constructing the [ImageAnalysisResultDetails](/java/api/azure.ai.vision.imageanalysis.imageanalysisresultdetails) object.
140
+
1. If succeeded, proceed to call the relevant result methods based on your selected visual features, as shown here. Additional information (not commonly needed) can be obtained by constructing the [ImageAnalysisResultDetails](/java/api/azure.ai.vision.imageanalysis.imageanalysisresultdetails) object.
141
141
142
142
1. If failed, you can construct the [ImageAnalysisErrorDetails](/java/api/azure.ai.vision.imageanalysis.imageanalysisresultdetails) object to get information on the failure.
143
143
@@ -149,7 +149,7 @@ This section shows you how to make an analysis call to the service using the sta
149
149
This section shows you how to make an analysis call to the service, when using a custom model.
150
150
151
151
152
-
The code is similar to the standard model case. The only difference is that results from the custom model are available on by calling **getCustomTags** and/or **getCustomObjects** methods on the [ImageAnalysisResult](/java/api/azure.ai.vision.imageanalysis.imageanalysisresult) object.
152
+
The code is similar to the standard model case. The only difference is that results from the custom model are available by calling **getCustomTags** and/or **getCustomObjects** methods on the [ImageAnalysisResult](/java/api/azure.ai.vision.imageanalysis.imageanalysisresult) object.
0 commit comments