The documentation is still under active development. For now we have the following resources available:
- Create a new Supervisely plugin
- Easy guide: Integrate a custom Pytorch Segmentation neural network
- General detailed guide: Integrate any custom neural network
- Deploy neural network as API
- Develop NN plugin
- How to attach model trained outside Supervisely (for TF object detection plugin)
- Working with Supervisely projects and labeling data
- Scripting interactions with the web instance using Supervisely API
- Data augmentation for neural network training
- Neral network inference and online deployment using Supervisely API
- Automating comple workflows using Supervisely API
- Explanation of different inference modes for NN: full image, sliding window, roi
- How to copy, move and delete data using py-SDK and REST-API
- How to manage users and labeling jobs
- Custom inference pipeline: image -> detection -> segmentation -> postprocessing
- Custom upload procedure: check if image exists and how to upload only new images to Supervisely instance
- Custom data pipeline: upload -> auto labeling jobs -> move labeled data to "secret" project
- Filter images in different projects and combine results into a new project
- How to apply NN (UNet/YOLO/Mask-RCNN) to the image from sources
- Apply NN to image parts (slidign window inference)
- Analyse data annotation quality
- Calculate classification metrics
- Calculate confusion matrix
- Calculate mean average precision (mAP)
- Calculate mean intersection over union (mIOU)
- Convert between class geometry types
- Import a project using a list of image links
- Download a project locally
- Filter project by tags
- Merge projects into one
- Plot tags distribution statistics
- Split the data between train and validation folds using tags
- Add augmentations and prepare data for training a detection model
- Add augmentations and prepare data for training a segmentation model
- Upload a project using using Supervisely API