Generic backend for interactive feature-based segmentation in python.
To install:
git clone https://github.com/tldr-group/interactive-seg-backend
cd interactive-seg-backend
pip install .
For GPU-enabled featurising (recommended), install with:
pip install '.[gpu]'
For development (linters, tests), install with
pip install -e '.[lint,test]'
To get all the optional dependencies at once:
pip install '.[all]'
python -m cProfile -s tottime multiscale_classical_cpu.py > bench.txt
Requires the pytest package (pip install '.[test]'
)
pytest -s
pip uninstall interactive_seg_backend -y
pip install -e . --no-cache-dir
If offline
pip uninstall interactive_seg_backend -y
pip install . --no-cache-dir --no-index
-
applying: patched, 3D (+ average), all with memory consideration (caching)
-
improvements: fixed vf, autocontext, rules?, crf?
-
make CPU version still work with conditional imports and string quote types
-
make typing story more compelling:
- actually make the main functions able to take in tensors or arrays
- make core vs main distinction make more sense
- export everything important from main / init
-
docstrings