Slideflow DocumentationΒΆ
Slideflow is a Python package that provides a unified API for building and testing deep learning models for histopathology, supporting both Tensorflow/Keras and PyTorch.
Slideflow includes tools for efficient whole-slide image processing, easy and highly customizable model training with uncertainty quantification (UQ), and a number of functional tools to assist with analysis and interpretability, including predictive heatmaps, mosaic maps, GANs, saliency maps, and more. It is built with both Tensorflow/Keras and PyTorch backends, with fully cross-compatible TFRecord data storage.
This documentation starts with a high-level overview of the pipeline and includes examples of how to perform common tasks using the Project
helper class. We also provide several tutorials with examples of how Slideflow can be used and extended for additional functionality.
- Installation
- Overview
- Quickstart
- Setting up a Project
- Datasets
- Slide Processing
- Training
- Evaluation
- Layer Activations
- Uncertainty Quantification
- Generating Features
- Multiple-Instance Learning (MIL)
- Self-Supervised Learning (SSL)
- Generative Networks (GANs)
- Saliency Maps
- Tissue Segmentation
- Cell Segmentation
- Custom Training Loops
- Slideflow Studio: Live Visualization
- Troubleshooting
- slideflow
- slideflow.Project
- slideflow.Dataset
- slideflow.DatasetFeatures
- slideflow.Heatmap
- slideflow.ModelParams
- slideflow.Mosaic
- slideflow.SlideMap
- slideflow.biscuit
- slideflow.cellseg
- slideflow.io
- slideflow.io.tensorflow
- slideflow.io.torch
- slideflow.gan
- slideflow.grad
- slideflow.mil
- slideflow.model
- slideflow.model.tensorflow
- slideflow.model.torch
- slideflow.norm
- slideflow.simclr
- slideflow.slide
- slideflow.slide.qc
- slideflow.stats
- slideflow.util
- slideflow.studio
- Tutorial 1: Model training (simple)
- Tutorial 2: Model training (advanced)
- Tutorial 3: Using a custom architecture
- Tutorial 4: Model evaluation & heatmaps
- Tutorial 5: Creating a mosaic map
- Tutorial 6: Custom slide filtering
- Tutorial 7: Training with custom augmentations
- Tutorial 8: Multiple-Instance Learning