/annot3d

Interactive annotation tool for 3D TIFF volumes with planar views, integrated automatic UNET predictions and live volume rendering for semi-automated semantic binary annotations.

Primary LanguagePython

3D Annotation Tool for TIFF volumes with Live UNET Segmentation

Interactive annotation tool for 3D TIFF volumes with integrated UNET semantic segmentation for semi-automated and automated binary annotations along with embedded live volume rendering. The tool offers planar views into all cross-sections and establishes correspondences between them while storing a 3D representation of the binary annotation voxel grid. pyside2_gui

Dependencies:

Install all required dependencies by:

pip install -r requirements.txt

Run:

python main.py

Future prospects and development:

  • Active deep learning for semi-automated/automated annotation with integrated Keras/TF training utilization for GPU-powered desktops. (WIP)
  • UNET diverse modified segmentation models and other auto-encoder alternatives as well as dynamic model specifications.
  • Data augmentation setup and adjustments using elastic deformations.
  • Export of the minimal annotation features to Android/iOS with access to a render/training server.
  • Improved and smarter real-time volume rendering using a VTK pipeline.
  • Graphics Tablet for Desktop & Natural User Interface (NUI) for interactions on Mobile/Tablet devices and usability.
  • CUDA core processing for numpy via CuPy or OpenCUDA integration.
  • Registration of 3D segmented datasets and annotations.
  • Session tracking, autosave, more utility features and multi-class segmentation.

History:

V0.2 Kivy GUI

Shifted to Kivy for a cross-platform (Desktop/Mobile) and responsive GUI experience with pure python libraries for export via python-for-android and utilization of OpenGL. kivy_gui

V0.1 Tkinter GUI

Implemented with Tkinter initially with popups for maximizing plane views and restricted window sizes along with the first functions. tkinter_gui