1.12.5
Python SDK
Superb AI Suite SDK Sample
Superb AI Suite SDK is available via Github. Please refer to the following link for more guides and example SDK use-cases: Superb AI Suite Examples SDK
- Image Basic: Basic SDK Example for Image projects
- Polygon to Bitmask: Convert exported polygon segmentation result to bitmask
- Suite-Pachyderm Integration: Suite - Pachyderm import / export examples
Initialization
First, follow the steps in the guide below to install and configure the CLI-SDK package.
Initialize the Suite SDK
Once you're configured, you can initialize the Suite Python SDK as below.
1
import spb.sdk
2
3
# Print Suite SDK version
4
print(spb.sdk.__version__)
Copied!
Initialize the SDK Client to access your project
List all projects (using paginated API)
1
client = spb.sdk.Client()
2
3
total_project_count, projects = client.get_projects(page=1, page_size=10)
4
for project in projects:
5
print('Project Name: {}'.format(project.name))
Copied!
Access project given project name
1
# Set Project using set_project
2
project = client.get_project('<Your Project Name>')
3
client.set_project(project)
4
5
# Print project information
6
print('Project Name: {}'.format(client.get_project_name()))
7
print('Total number of data: {}'.format(client.get_num_data()))
Copied!
Upload An Image to Suite
Upload a local image
1
image_path = '<Local Path to Your Image>'
2
dataset = '<Your Dataset Name>'
3
4
client.upload_image(image_path, dataset)
5
6
# The uploaded image can be seen at the Suite Web UI
Copied!
Upload an image from your Amazon S3 bucket
You need AWS credentials to access your S3 bucket.
1
bucket = '<Your S3 Bucket Name>'
2
image_path = '<Path to your Image in the Bucket>'
3
dataset = '<Your Dataset Name>'
4
5
client.upload_image_s3(bucket, image_path, dataset)
6
7
# The uploaded image can be seen at the Suite Web UI
Copied!
Access Raw Data on Suite
Get a data row from Suite
1
# Generate iterator from selected project
2
data_iter = client.get_data_page(page_idx='<Index of pages>', page_size='<Number of data in page>')
3
data = next(data_iter)
4
5
# Print data information
6
print('Key: {}'.format(data.get_key()))
7
print('Dataset: {}'.format(data.get_dataset_name()))
8
print('Image URL: {}'.format(data.get_image_url()))
9
print('Status: {}'.format(data.get_status()))
Copied!
Read the corresponding image from a data row
1
image = data.get_image()
2
3
# Print image information
4
print('Image type: {}'.format(image.__class__))
5
print('Image dtype: {}'.format(image.dtype))
6
print('Image shape: {}'.format(image.shape))
7
8
# Plot the image
9
import matplotlib.pyplot as plt
10
plt.imshow(image)
Copied!
Download the image file to local storage.
1
download_path, result = data.download_image(download_to='./test.jpg')
2
3
# Print the HTTP response
4
print(result)
5
6
# Plot the downloaded image
7
import PIL.Image
8
image = PIL.Image.open(download_path)
9
plt.imshow(image)
Copied!
Access Raw Data with Filters
Filtering via Dataset
1
dataset = 'MSCOCO_dataset_a'
2
3
# Total number of data in your dataset
4
num_data = client.get_num_data(dataset=dataset)
5
print('Size of "{}" dataset: {}'.format(dataset, num_data))
6
7
# Get data iterator in the dataset
8
data_iter = client.get_data_page(page_idx='<Index of pages>', page_size='<Number of data in page>', dataset=dataset)
9
data = next(data_iter)
Copied!
Filtering via Label Tags
1
tags = ['<Label Tag 1>', '<Label Tag 2>', ...]
2
3
# Number of data containing all the tags
4
num_data = client.get_num_data(tags=tags)
5
print('Number of data containing tags {}: {}'.format(tags, num_data))
6
7
# Get data iterator in the filtered list of data
8
data_iter = client.get_data_page(page_idx='<Index of pages>', page_size='<Number of data in page>', tags=tags)
9
data = next(data_iter)
10
11
# Print tags of the data
12
print('Tags: {}'.format(data.get_tags()))
Copied!
Manage labels on Suite
Get the label information for each data row
1
category_labels = data.get_category_labels()
2
object_labels = data.get_object_labels()
3
4
# Print labels
5
print('category_labels', category_labels)
6
print('object_labels', object_labels)
7
# Example output:
8
# category_labels ['bugatti']
9
# object_labels [
10
# {'id': 1, 'class': 'bugatti',
11
# 'shape': {'box': {'x': 83, 'y': 52, 'width': 819, 'height': 457}},
12
# 'properties': []},
13
# {'id': 2, 'class': ...},
14
# {'id': 3, 'class': ...},
15
# ...
16
# ]
Copied!
Change the label information locally
1
category_labels = ['teslaroadster']
2
object_labels[0]['class'] = 'teslaroadster'
3
4
# Print labels
5
print('category_labels', category_labels)
6
print('object_labels', object_labels)
7
# Example output:
8
# category_labels ['teslaroadster']
9
# object_labels [
10
# {'id': 1, 'class': 'teslaroadster',
11
# 'shape': {'box': {'x': 83, 'y': 52, 'width': 819, 'height': 457}},
12
# 'properties': []},
13
# {'id': 2, 'class': ...},
14
# {'id': 3, 'class': ...},
15
# ...
16
# ]
Copied!
Apply the changes to Suite (update label information)
1
data.set_category_labels(category_labels)
2
data.set_object_labels(object_labels)
3
4
# The change can be seen at the Suite Web UI
Copied!
Last modified 1mo ago
Copy link
Contents
Superb AI Suite SDK Sample
Initialization
Initialize the Suite SDK
Initialize the SDK Client to access your project
Upload An Image to Suite
Upload a local image
Upload an image from your Amazon S3 bucket
Access Raw Data on Suite
Get a data row from Suite
Read the corresponding image from a data row
Download the image file to local storage.
Access Raw Data with Filters
Filtering via Dataset
Filtering via Label Tags
Manage labels on Suite
Get the label information for each data row
Change the label information locally
Apply the changes to Suite (update label information)