ibm_watson.visual_recognition_v3 module¶
The IBM Watson™ Visual Recognition service uses deep learning algorithms to identify scenes and objects in images that you upload to the service. You can create and train a custom classifier to identify subjects that suit your needs.
-
class
VisualRecognitionV3(version, authenticator=None)[source]¶ Bases:
ibm_cloud_sdk_core.base_service.BaseServiceThe Visual Recognition V3 service.
-
default_service_url= 'https://gateway.watsonplatform.net/visual-recognition/api'¶
-
classify(*, images_file=None, images_filename=None, images_file_content_type=None, url=None, threshold=None, owners=None, classifier_ids=None, accept_language=None, **kwargs)[source]¶ Classify images.
Classify images with built-in or custom classifiers.
- Parameters
images_file (file) – (optional) An image file (.gif, .jpg, .png, .tif) or .zip file with images. Maximum image size is 10 MB. Include no more than 20 images and limit the .zip file to 100 MB. Encode the image and .zip file names in UTF-8 if they contain non-ASCII characters. The service assumes UTF-8 encoding if it encounters non-ASCII characters. You can also include an image with the url parameter.
images_filename (str) – (optional) The filename for images_file.
images_file_content_type (str) – (optional) The content type of images_file.
url (str) – (optional) The URL of an image (.gif, .jpg, .png, .tif) to analyze. The minimum recommended pixel density is 32X32 pixels, but the service tends to perform better with images that are at least 224 x 224 pixels. The maximum image size is 10 MB. You can also include images with the images_file parameter.
threshold (float) – (optional) The minimum score a class must have to be displayed in the response. Set the threshold to 0.0 to return all identified classes.
owners (list[str]) – (optional) The categories of classifiers to apply. The classifier_ids parameter overrides owners, so make sure that classifier_ids is empty. - Use IBM to classify against the default general classifier. You get the same result if both classifier_ids and owners parameters are empty. - Use me to classify against all your custom classifiers. However, for better performance use classifier_ids to specify the specific custom classifiers to apply. - Use both IBM and me to analyze the image against both classifier categories.
classifier_ids (list[str]) – (optional) Which classifiers to apply. Overrides the owners parameter. You can specify both custom and built-in classifier IDs. The built-in default classifier is used if both classifier_ids and owners parameters are empty. The following built-in classifier IDs require no training: - default: Returns classes from thousands of general tags. - food: Enhances specificity and accuracy for images of food items. - explicit: Evaluates whether the image might be pornographic.
accept_language (str) – (optional) The desired language of parts of the response. See the response for details.
headers (dict) – A dict containing the request headers
- Returns
A DetailedResponse containing the result, headers and HTTP status code.
- Return type
DetailedResponse
-
create_classifier(name, positive_examples, *, negative_examples=None, negative_examples_filename=None, **kwargs)[source]¶ Create a classifier.
Train a new multi-faceted classifier on the uploaded image data. Create your custom classifier with positive or negative example training images. Include at least two sets of examples, either two positive example files or one positive and one negative file. You can upload a maximum of 256 MB per call. Tips when creating: - If you set the X-Watson-Learning-Opt-Out header parameter to true when you create a classifier, the example training images are not stored. Save your training images locally. For more information, see [Data collection](#data-collection). - Encode all names in UTF-8 if they contain non-ASCII characters (.zip and image file names, and classifier and class names). The service assumes UTF-8 encoding if it encounters non-ASCII characters.
- Parameters
name (str) – The name of the new classifier. Encode special characters in UTF-8.
positive_examples (dict) – A dictionary that contains the value for each classname. The value is a .zip file of images that depict the visual subject of a class in the new classifier. You can include more than one positive example file in a call. Specify the parameter name by appending _positive_examples to the class name. For example, goldenretriever_positive_examples creates the class goldenretriever. Include at least 10 images in .jpg or .png format. The minimum recommended image resolution is 32X32 pixels. The maximum number of images is 10,000 images or 100 MB per .zip file. Encode special characters in the file name in UTF-8.
negative_examples (file) – (optional) A .zip file of images that do not depict the visual subject of any of the classes of the new classifier. Must contain a minimum of 10 images. Encode special characters in the file name in UTF-8.
negative_examples_filename (str) – (optional) The filename for negative_examples.
headers (dict) – A dict containing the request headers
- Returns
A DetailedResponse containing the result, headers and HTTP status code.
- Return type
DetailedResponse
-
list_classifiers(*, verbose=None, **kwargs)[source]¶ Retrieve a list of classifiers.
- Parameters
- Returns
A DetailedResponse containing the result, headers and HTTP status code.
- Return type
DetailedResponse
-
get_classifier(classifier_id, **kwargs)[source]¶ Retrieve classifier details.
Retrieve information about a custom classifier.
-
update_classifier(classifier_id, *, positive_examples={}, negative_examples=None, negative_examples_filename=None, **kwargs)[source]¶ Update a classifier.
Update a custom classifier by adding new positive or negative classes or by adding new images to existing classes. You must supply at least one set of positive or negative examples. For details, see [Updating custom classifiers](https://cloud.ibm.com/docs/services/visual-recognition?topic=visual-recognition-customizing#updating-custom-classifiers). Encode all names in UTF-8 if they contain non-ASCII characters (.zip and image file names, and classifier and class names). The service assumes UTF-8 encoding if it encounters non-ASCII characters. Tips about retraining: - You can’t update the classifier if the X-Watson-Learning-Opt-Out header parameter was set to true when the classifier was created. Training images are not stored in that case. Instead, create another classifier. For more information, see [Data collection](#data-collection). - Don’t make retraining calls on a classifier until the status is ready. When you submit retraining requests in parallel, the last request overwrites the previous requests. The retrained property shows the last time the classifier retraining finished.
- Parameters
classifier_id (str) – The ID of the classifier.
positive_examples (dict) – (optional) A dictionary that contains the value for each classname. The value is a .zip file of images that depict the visual subject of a class in the classifier. The positive examples create or update classes in the classifier. You can include more than one positive example file in a call. Specify the parameter name by appending _positive_examples to the class name. For example, goldenretriever_positive_examples creates the class goldenretriever. Include at least 10 images in .jpg or .png format. The minimum recommended image resolution is 32X32 pixels. The maximum number of images is 10,000 images or 100 MB per .zip file. Encode special characters in the file name in UTF-8.
negative_examples (file) – (optional) A .zip file of images that do not depict the visual subject of any of the classes of the new classifier. Must contain a minimum of 10 images. Encode special characters in the file name in UTF-8.
negative_examples_filename (str) – (optional) The filename for negative_examples.
headers (dict) – A dict containing the request headers
- Returns
A DetailedResponse containing the result, headers and HTTP status code.
- Return type
DetailedResponse
-
get_core_ml_model(classifier_id, **kwargs)[source]¶ Retrieve a Core ML model of a classifier.
Download a Core ML model file (.mlmodel) of a custom classifier that returns <tt>”core_ml_enabled”: true</tt> in the classifier details.
-
delete_user_data(customer_id, **kwargs)[source]¶ Delete labeled data.
Deletes all data associated with a specified customer ID. The method has no effect if no data is associated with the customer ID. You associate a customer ID with data by passing the X-Watson-Metadata header with a request that passes data. For more information about personal data and customer IDs, see [Information security](https://cloud.ibm.com/docs/services/visual-recognition?topic=visual-recognition-information-security).
-
-
class
Class(class_)[source]¶ Bases:
objectA category within a classifier.
- Attr str classclass_
The name of the class.
-
class
ClassResult(class_, score, *, type_hierarchy=None)[source]¶ Bases:
objectResult of a class within a classifier.
- Attr str classclass_
Name of the class. Class names are translated in the language defined by the Accept-Language request header for the build-in classifier IDs (default, food, and explicit). Class names of custom classifiers are not translated. The response might not be in the specified language when the requested language is not supported or when there is no translation for the class name.
- Attr float score
Confidence score for the property in the range of 0 to 1. A higher score indicates greater likelihood that the class is depicted in the image. The default threshold for returning scores from a classifier is 0.5.
- Attr str type_hierarchy
(optional) Knowledge graph of the property. For example, /fruit/pome/apple/eating apple/Granny Smith. Included only if identified.
-
class
ClassifiedImage(classifiers, *, source_url=None, resolved_url=None, image=None, error=None)[source]¶ Bases:
objectResults for one image.
- Attr str source_url
(optional) Source of the image before any redirects. Not returned when the image is uploaded.
- Attr str resolved_url
(optional) Fully resolved URL of the image after redirects are followed. Not returned when the image is uploaded.
- Attr str image
(optional) Relative path of the image file if uploaded directly. Not returned when the image is passed by URL.
- Attr ErrorInfo error
(optional) Information about what might have caused a failure, such as an image that is too large. Not returned when there is no error.
- Attr list[ClassifierResult] classifiers
The classifiers.
-
class
ClassifiedImages(images, *, custom_classes=None, images_processed=None, warnings=None)[source]¶ Bases:
objectResults for all images.
- Attr int custom_classes
(optional) Number of custom classes identified in the images.
- Attr int images_processed
(optional) Number of images processed for the API call.
- Attr list[ClassifiedImage] images
Classified images.
- Attr list[WarningInfo] warnings
(optional) Information about what might cause less than optimal output. For example, a request sent with a corrupt .zip file and a list of image URLs will still complete, but does not return the expected output. Not returned when there is no warning.
-
class
Classifier(classifier_id, name, *, owner=None, status=None, core_ml_enabled=None, explanation=None, created=None, classes=None, retrained=None, updated=None)[source]¶ Bases:
objectInformation about a classifier.
- Attr str classifier_id
ID of a classifier identified in the image.
- Attr str name
Name of the classifier.
- Attr str owner
(optional) Unique ID of the account who owns the classifier. Might not be returned by some requests.
- Attr str status
(optional) Training status of classifier.
- Attr bool core_ml_enabled
(optional) Whether the classifier can be downloaded as a Core ML model after the training status is ready.
- Attr str explanation
(optional) If classifier training has failed, this field might explain why.
- Attr datetime created
(optional) Date and time in Coordinated Universal Time (UTC) that the classifier was created.
- Attr list[Class] classes
(optional) Classes that define a classifier.
- Attr datetime retrained
(optional) Date and time in Coordinated Universal Time (UTC) that the classifier was updated. Might not be returned by some requests. Identical to updated and retained for backward compatibility.
- Attr datetime updated
(optional) Date and time in Coordinated Universal Time (UTC) that the classifier was most recently updated. The field matches either retrained or created. Might not be returned by some requests.
-
class
ClassifierResult(name, classifier_id, classes)[source]¶ Bases:
objectClassifier and score combination.
- Attr str name
Name of the classifier.
- Attr str classifier_id
ID of a classifier identified in the image.
- Attr list[ClassResult] classes
Classes within the classifier.
-
class
Classifiers(classifiers)[source]¶ Bases:
objectA container for the list of classifiers.
- Attr list[Classifier] classifiers
List of classifiers.
-
class
ErrorInfo(code, description, error_id)[source]¶ Bases:
objectInformation about what might have caused a failure, such as an image that is too large. Not returned when there is no error.
- Attr int code
HTTP status code.
- Attr str description
Human-readable error description. For example, File size limit exceeded.
- Attr str error_id
Codified error string. For example, limit_exceeded.