【GCP】Vision API Product Search触ってみた。

1. オンラインで個別にインポートを行い、商品セットを作成する。

空の商品セットを作成する。これは商品のグループを格納するためのシンプルなコンテナです。

# def create_product_set(
#         project_id, location, product_set_id, product_set_display_name):
#     “””Create a product set.
#     Args:
#         project_id: Id of the project.
#         location: A compute region name.
#         product_set_id: Id of the product set.
#         product_set_display_name: Display name of the product set.
#     “””
#     client = vision.ProductSearchClient()
#     # A resource that represents Google Cloud Platform location.
#     location_path = f”projects/{project_id}/locations/{location}”
#     # Create a product set with the product set specification in the region.
#     product_set = vision.ProductSet(
#             display_name=product_set_display_name)
#     # The response is the product set with `name` populated.
#     response = client.create_product_set(
#         parent=location_path,
#         product_set=product_set,
#         product_set_id=product_set_id)
#     # Display the product set information.
#     print(‘Product set name: {}’.format(response.name))

2. 個別の商品を作成する。

商品セットを作成したら、次のステップとして商品を作成します。

# def create_product(
#         project_id, location, product_id, product_display_name,
#         product_category):
#     “””Create one product.
#     Args:
#         project_id: Id of the project.
#         location: A compute region name.
#         product_id: Id of the product.
#         product_display_name: Display name of the product.
#         product_category: Category of the product.
#     “””
#     client = vision.ProductSearchClient()
#     # A resource that represents Google Cloud Platform location.
#     location_path = f”projects/{project_id}/locations/{location}”
#     # Create a product with the product specification in the region.
#     # Set product display name and product category.
#     product = vision.Product(
#         display_name=product_display_name,
#         product_category=product_category)
#     # The response is the product with the `name` field populated.
#     response = client.create_product(
#         parent=location_path,
#         product=product,
#         product_id=product_id)
#     # Display the product information.
#     print(‘Product name: {}’.format(response.name))

3. 商品セットに商品を追加する。

商品セットと商品が作成されたら、商品セットに商品を追加できます。

# def add_product_to_product_set(
#         project_id, location, product_id, product_set_id):
#     “””Add a product to a product set.
#     Args:
#         project_id: Id of the project.
#         location: A compute region name.
#         product_id: Id of the product.
#         product_set_id: Id of the product set.
#     “””
#     client = vision.ProductSearchClient()
#     # Get the full path of the product set.
#     product_set_path = client.product_set_path(
#         project=project_id, location=location,
#         product_set=product_set_id)
#     # Get the full path of the product.
#     product_path = client.product_path(
#         project=project_id, location=location, product=product_id)
#     # Add the product to the product set.
#     client.add_product_to_product_set(
#         name=product_set_path, product=product_path)
#     print(‘Product added to product set.’)

作成した商品や商品セットを更新んする必要がある場合は、更新用のメソッドを使用できます。

# def update_product_labels(
#         project_id, location, product_id, key, value):
#     “””Update the product labels.
#     Args:
#         project_id: Id of the project.
#         location: A compute region name.
#         product_id: Id of the product.
#         key: The key of the label.
#         value: The value of the label.
#     “””
#     client = vision.ProductSearchClient()
#     # Get the name of the product.
#     product_path = client.product_path(
#         project=project_id, location=location, product=product_id)
#     # Set product name, product label and product display name.
#     # Multiple labels are also supported.
#     key_value = vision.Product.KeyValue(key=key, value=value)
#     product = vision.Product(
#         name=product_path,
#         product_labels=[key_value])
#     # Updating only the product_labels field here.
#     update_mask = field_mask.FieldMask(paths=[‘product_labels’])
#     # This overwrites the product_labels.
#     updated_product = client.update_product(
#         product=product, update_mask=update_mask)
#     # Display the updated product information.
#     print(‘Product name: {}’.format(updated_product.name))
#     print(‘Updated product labels: {}’.format(product.product_labels))

個別の商品の参照画像を作成すると、画像にインデックスが追加された後にVision API Product Searchでその画像を使用して商品を検索できます。

# def create_reference_image(
#         project_id, location, product_id, reference_image_id, gcs_uri):
#     “””Create a reference image.
#     Args:
#         project_id: Id of the project.
#         location: A compute region name.
#         product_id: Id of the product.
#         reference_image_id: Id of the reference image.
#         gcs_uri: Google Cloud Storage path of the input image.
#     “””
#     client = vision.ProductSearchClient()
#     # Get the full path of the product.
#     product_path = client.product_path(
#         project=project_id, location=location, product=product_id)
#     # Create a reference image.
#     reference_image = vision.ReferenceImage(uri=gcs_uri)
#     # The response is the reference image with `name` populated.
#     image = client.create_reference_image(
#         parent=product_path,
#         reference_image=reference_image,
#         reference_image_id=reference_image_id)
#     # Display the reference image information.
#     print(‘Reference image name: {}’.format(image.name))
#     print(‘Reference image uri: {}’.format(image.uri))

6. 類似商品を検索する。

新しい画像を入力として受け取り、最も一致する商品を検索することで、作成した商品カタログを区クエリ出来ます。

# def get_similar_products_file(
#         project_id,
#         location,
#         product_set_id,
#         product_category,
#         file_path,
#         filter,
#         max_results
# ):
#     “””Search similar products to image.
#     Args:
#         project_id: Id of the project.
#         location: A compute region name.
#         product_set_id: Id of the product set.
#         product_category: Category of the product.
#         file_path: Local file path of the image to be searched.
#         filter: Condition to be applied on the labels.
#                 Example for filter: (color = red OR color = blue) AND style = kids
#                 It will search on all products with the following labels:
#                 color:red AND style:kids
#                 color:blue AND style:kids
#         max_results: The maximum number of results (matches) to return. If omitted, all results are returned.
#     “””
#     # product_search_client is needed only for its helper methods.
#     product_search_client = vision.ProductSearchClient()
#     image_annotator_client = vision.ImageAnnotatorClient()

#     # Read the image as a stream of bytes.
#     with open(file_path, ‘rb’) as image_file:
#         content = image_file.read()

#     # Create annotate image request along with product search feature.
#     image = vision.Image(content=content)

#     # product search specific parameters
#     product_set_path = product_search_client.product_set_path(
#         project=project_id, location=location,
#         product_set=product_set_id)
#     product_search_params = vision.ProductSearchParams(
#         product_set=product_set_path,
#         product_categories=[product_category],
#         filter=filter)
#     image_context = vision.ImageContext(
#         product_search_params=product_search_params)

#     # Search products similar to the image.
#     response = image_annotator_client.product_search(
#         image,
#         image_context=image_context,
#         max_results=max_results
#     )
#     index_time = response.product_search_results.index_time
#     print(‘Product set index time: ‘)
#     print(index_time)

#     results = response.product_search_results.results

#     print(‘Search results:’)
#     for result in results:
#         product = result.product
#         print(‘Score(Confidence): {}’.format(result.score))
#         print(‘Image name: {}’.format(result.image))
#         print(‘Product name: {}’.format(product.name))
#         print(‘Product display name: {}’.format(
#             product.display_name))
#         print(‘Product description: {}\n’.format(product.description))
#         print(‘Product labels: {}\n’.format(product.product_labels))