pyEdgeEval.helpers package

Submodules

pyEdgeEval.helpers.convert_cityscapes module

Convert Cityscapes to SBD format

TODO: - each process uses around 10GB (potentially OOM on constrained systems) - debug memory usage and unnecessary allocations (probably cv2)

pyEdgeEval.helpers.convert_cityscapes.convert_cityscapes(proc_dirname: str, cityscapes_root: str = 'data/cityscapes', inst_sensitive: bool = True, ext: str = 'png', save_half_val: bool = True, train_radius: int = 2, raw_val_radius: int = 2, thin_val_radius: int = 1, nproc: int = 4)[source]
pyEdgeEval.helpers.convert_cityscapes.test_against_matlab(proc_dirname: str, matlab_proc_dirname: str = 'gtProc', cityscapes_root: str = 'data/cityscapes', inst_sensitive: bool = True, ext: str = 'png', nproc: int = 4)[source]

pyEdgeEval.helpers.evaluate_bsds500 module

pyEdgeEval.helpers.evaluate_bsds500.evaluate_bsds500(no_split_dir=False)[source]

pyEdgeEval.helpers.evaluate_cityscapes module

General evaluation protocols for Cityscapes Semantic Boundary Benchmark

The ‘Raw’ Protocol: - Use CityscapesEvaluator or HalfcityscapesEvaluator

The ‘Thin’ Protocol: - Use HalfCityscapesEvaluator

pyEdgeEval.helpers.evaluate_cityscapes.evaluate_cityscapes_thin(gt_dir: str = 'gtEval')[source]
pyEdgeEval.helpers.evaluate_cityscapes.evaluate_cityscapes_raw(gt_dir: str = 'gtEval')[source]

pyEdgeEval.helpers.evaluate_sbd module

pyEdgeEval.helpers.evaluate_sbd.evaluate_sbd()[source]

Module contents