Supervisely SDK for Python

Python SDK Reference:

• API Reference

Public REST API Reference:

• Public REST API

Guides:

• Intro
• What is Supervisely Agent
• Custom plugin basics
• Custom NN plugin basics
• How to debug custom NN plugin
• Different ways how to deploy NN with Supervisely
• Exam reports explained
• UI documentation

Basics and Usage Examples:

• Project Structure
  • Supervisely Python SDK Tutorial #1
  • Projects and datasets basics
    • Hello Project - explore the data
      • Image annotations and meta-information
      • Geometric labels
      • Image-level tags
    • Create and populate a new project
      • Create a new empty project
      • Set up new project metadata
      • Populate the new project with labeled images
      • Direct on-disk project copy
    • Modifying annotations and labels
      • Built-in transformations
      • Construct new labels externally
      • Label transform functions
    • Working with tags
      • No-value tags
      • Tags with values
• Data Management
  • Supervisely Tutorial #2
  • Online API basics: organize and explore workspaces, projects and neural networks
    • Necessary imports
      • Initialize API access with your credentials
      • Workspace management
      • Project management
      • Neural network management
• Augmentations
  • Supervisely Python SDK Tutorial #3: Augmentations
  • Imports
    • Required imports
      • Prepare sample input
    • Load input data
    • Define visualization helpers
    • Original image and its annotation
      • Spatial augmentations
    • Horizontal flip (left-right)
    • Vertical flip
    • Random crop - height and width from range
    • Random crop - fixed height and width
    • Deterministic crop (relative units)
    • Deterministic crop (absolute pixels)
    • Resize
    • Scale proportionally
    • Rotate (with padding)
    • Rotate (without padding)
      • Per-pixel augmentations
    • Random brightness
    • Random contrast
    • Random Gaussian noise
    • Random per-color scale
    • Blur
    • Median Blur
    • Gaussian Blur
• NN: deploy and inference on Supervisely via API
  • Supervisely Tutorial #4
  • Neural networks: deploy and inference with Supervisely online API
    • Setup steps
      • Necessary imports
      • Initialize API access with your credentials
      • Define the active workspace
      • Add the neural network model to the workspace
      • Select the agent to use
    • Online on-demand inference
      • Deploy the model to the agent for on-demand inference
      • Get the metadata for the deployed model
      • Inference with a locally stored image
      • Inference with an external image via web link
      • Inference with images from Supervisely projects
    • Stop the deployed model
• Automate NN training and inference via API
  • Supervisely Tutorial #5
  • Neural networks: training workflow with Supervisely online API
    • Setup steps
      • Necessary imports
      • Initialize API access with your credentials
      • Define the active workspace
    • Add a neural network to the workspace to be a starting point for training
      • Select the agent to use
    • Prepare a project with the training data
      • Copy an existing labeled project
      • Data augmentation using a Data Transformation Language plugin
    • Run neural network training
    • Run inference on the freshly trained model
      • Prepare an input dataset for inference
      • Run inference on the input project
    • Done!
• Inference modes: full image / sliding window / ROI / bboxes
  • Inference modes
    • Inference config structure
    • Full image inference
    • ROI (fixed crop) inference
    • Sliding window inference
      • Segmentation
      • Detection
    • Pipelining - bounding boxes ROI inference
• Data manipulation via API: copy / move / delete
  • Imports
  • Initialize API access with your credentials
  • Script parameters
  • Verify parameters
  • Clone demo project from Explore page
  • Helper functions
  • Use cases: copy/move/delete
    • Print original project structure
    • Copy dataset
    • Copy images from one dataset to another
    • Move some images to new dataset
    • Delete images from dataset
    • Remove datasets from project
    • Remove entire demo project
• Users and labeling jobs API
  • Imports
  • Initialize API access with your credentials
  • Print all roles that are available on private Supervisely instance
  • List all registered users
  • Get UserInfo by ID
  • Get UserInfo by login
  • Update user info
  • Get User Membership (list all user teams with corresponding roles)
  • Create new user
  • Disable/Enable user
  • Invite user to team
  • List all team users with corresponding roles
  • Change user role in team
  • Remove user from team
  • Labeling jobs
    • Step 1. Before creating labeling Job, it is needed to add annotators to team
    • Step 2. Define project and datasets to label
    • Labeler1 will label cars on the first dataset
    • Stop Labeling Job, job will be unavailable for labeler
    • Labeler2 will label cars on the first dataset
  • Get all labeling jobs in a team
    • Labeling Jobs Filtering (filters [created_by_id, assigned_to_id, project_id, dataset_id] can be used in various combinations)
    • Get all labeling that were created by user ‘max’
    • Get all labeling that were created by user ‘max’ and were assigned to labeler02
    • Archive Labeling Job
    • Get all active labeling jobs in a team
  • Labeling Jobs Statuses
• Custom NN pilepile: (1) detect objects and (2) segment them with separate networks
  • Imports
  • Initialize API access with your credentials
  • Initialize the API access object
  • Script parameters
  • Verify parameters
  • Deploy both detection and segmentataion models (if not deployed before)
  • Get models ouptut classes (+ check that models are deployed correctly)
  • Apply detection model (sliding window)
  • Apply segmentation model (sliding window)
  • Postprocessing step: keep only ties that are on the cats
  • Mission complete!
• Upload images via API
  • Supervisely Tutorial #10
  • Imports
  • Initialize API access with your credentials
  • Initialize the API access object
  • Script parameters
  • Verify parameters
  • Find project or create the new one
  • Get paths of all images in a directory
  • Calculate hashes for all images paths
  • Get hashes that are already on server
  • Create dataset for new images
  • Upload images to the created dataset
  • Done!
• Custom Data Pipeline
  • Supervisely Tutorial #11
  • Imports
  • Initialize API access with your credentials
  • Initialize the API access object
  • Script parameters
  • Verify of initialize parameters
    • Populate projects “cats” and “dogs” with classes to label
  • Create new dataset for new batch of data
  • Upload new incoming data
  • Copy uploaded data to labeling projects
  • Create labeling jobs
    • Additional Info:
  • Combine labeled cats and docs and copy dataset to the final project
  • Done!
• Filter and combine images
  • Filter and combine images (custom scenario)
    • Imports
    • Initialize API access with your credentials
    • Initialize the API access object
    • Script parameters
    • Verify input values
    • Create resulting project
    • Iterate over all images, filter and combine them if needed
• How to work with NN pixelwise probabilities
  • NN pixelwise confidences: different inference modes, save and load.
    • Beginning
      • Imports
      • Initialize API access with your credentials
      • Script parameters
      • Get access to workspace/team/agent
      • Deploy model for segmentation.
    • Get all pixelwise scores for some image
      • Prepare inference config
      • Take some image…
      • … and run inference with the determined config
      • Explore results
    • Different inference modes
      • Full image
      • ROI (Region Of Interest)
      • Bounding Boxes
      • Sliding Window
    • Upload confidences to Supervisely server and download from server
      • Preparations
      • Create target project
      • Run inference and upload results
      • Download stored project and compare bitmaps with uploaded ones
      • Check the same work with other API method
    • Clean up: undeploy model
• Additional examples in our github

Indices and tables

• Index

• Module Index

• Search Page