Adds feedback examples to the dataset associated with the specified object detection model. Available in Einstein Vision API version 2.0 and later.
Request Parameters
Name | Type | Description | Available Version |
---|---|---|---|
| string | Local .zip file to upload. The maximum .zip file size you can upload from a local drive is 50 MB. The images and an annotations.csv file with the label and bounding box information are contained in a .zip file, just like when you create a dataset from a .zip file. | 2.0 |
| string | ID of the model that misclassified the images. The feedback examples are added to the dataset associated with this model. | 2.0 |
If a model returns an incorrect prediction for an image, you can use that image to improve the model. You do this by adding the image to the dataset and retraining it to update the model. The misclassified images that you add to the dataset are known as feedback. Use this API to add misclassified images with the correct labels to the dataset from which the model was created.
This call supports only datasets that have a type of image-detection
.
Keep the following points in mind when creating feedback examples.
-
You pass in a
modelId
parameter, but the examples are added to the dataset from which the specified model was created. -
The .zip file that contains the feedback images and annotations file follows the same format and structure as the .zip file you use to create a dataset. See the Object Detection Datasets section in Create a Dataset From a Zip File Asynchronously.
-
This API call checks for duplicate images in the .zip file that contains the feedback images. If the .zip file contains multiple image files that have the same contents, only the first of the duplicate files is uploaded.
-
The call also checks for duplicates between the images in the .zip file and the images already in the dataset. When the feedback examples and data from the annotations.csv file are merged, the API reconciles any differences by replacing earlier examples with the feedback images in the .zip file.
-
The maximum image file size is 5 MB.
-
Images must be no larger than 1,600 pixels high by 1,600 pixels wide. You can upload images that are larger, but training the dataset might fail.
-
The supported image file types are PNG, JPG, and JPEG.
-
This call is asynchronous, so the response has a status of
UPLOADING
. -
Use the dataset ID and make a call to Get a Dataset to query when the upload is complete. When available is true and statusMsg is
SUCCEEDED
the upload of feedback examples is complete.
Response Body
Name | Type | Description | Available Version |
---|---|---|---|
| boolean | Specifies whether the dataset is ready to be trained. | 2.0 |
| date | Date and time that the dataset was created. | 2.0 |
| long | ID of the dataset to which the feedback examples are added. | 2.0 |
| object | Contains the | 2.0 |
| string | Name of the dataset. | 2.0 |
| string | Object returned; in this case, | 2.0 |
| string | Status of the dataset while feedback is being added. Valid values are:
| 2.0 |
| int | Total number of examples in the dataset. | 2.0 |
| int | Total number of labels in the dataset. | 2.0 |
| string | Type of dataset data. Feedback examples can be added via a .zip file only for object detection datasets, so this returns | 2.0 |
| date | Date and time that the dataset was last updated. | 2.0 |
Label Response Body
Name | Type | Description | Available Version |
---|---|---|---|
| long | ID of the dataset to which the label belongs. | 2.0 |
| long | ID of the label. | 2.0 |
| string | Name of the label. | 2.0 |
| int | Number of examples that have the label. This is the number of examples before the feedback examples are added to the dataset. | 2.0 |