Open-source software for deep learning-based analysis and visualization of whole slide images in digital pathology.
FastPathology is an open-source software platform for deep learning-based analysis of whole slide images in digital pathology. It provides a graphical interface to deploy and run convolutional neural networks for tasks like segmentation, classification, and object detection on medical histopathology slides. The platform supports real-time visualization and multiple inference engines to accelerate model execution.
Digital pathology researchers, medical imaging scientists, and pathologists who need to apply deep learning models to whole slide images without extensive programming. It is also suitable for developers building tools for computational pathology.
FastPathology offers a code-free environment with a user-friendly GUI, making advanced deep learning accessible to non-programmers. Its support for multiple high-performance inference engines (TensorRT, OpenVINO, etc.) enables fast, real-time analysis of large whole slide images, which is critical in medical research and diagnostics.
⚡ Open-source software for deep learning-based digital pathology
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
Provides a user-friendly graphical interface that allows pathologists to deploy and run neural networks on whole slide images without any coding, as emphasized in the README's key features.
Supports various inference engines including TensorFlow, TensorRT, OpenVINO, and ONNX Runtime, enabling optimized performance for different hardware setups, detailed in the features section.
Offers low-memory cost streaming of predictions on top of WSIs in real-time, facilitating immediate feedback during analysis, as highlighted in the visualization feature.
Leverages OpenSlide to support a wide range of whole slide image formats, making it compatible with common pathology slide scanners, mentioned in the formats support.
Enabling high-speed NVIDIA GPU inference requires manual installation of specific versions of CUDA, cuDNN, and TensorRT, which can be error-prone and time-consuming, as noted in the optional setup section.
The macOS version is labeled as experimental and does not support NVIDIA GPU inference, restricting performance on Apple hardware, stated in the macOS installation notes.
Uninstallation involves manually deleting user folders for stored projects and cache, which could lead to leftover data and is less user-friendly than automated cleanup, as described in the installation instructions.