Shapely 1.8.5 lead to AttributeError: 'Polygon' object has no attribute 'uid'

[lydiayliu]

Using cellvit-inference with version 1.8.5 of Shapely resulted in the following error:

Traceback (most recent call last):
  File "/nemo/lab/swantonc/working/liul/.conda/envs/cellvit_env/bin/cellvit-inference", line 8, in <module>
    sys.exit(main())
  File "/nemo/lab/swantonc/working/liul/.conda/envs/cellvit_env/lib/python3.10/site-packages/cellvit/detect_cells.py", line 50, in main
    celldetector.process_wsi(
  File "/nemo/lab/swantonc/working/liul/.conda/envs/cellvit_env/lib/python3.10/site-packages/cellvit/inference/inference.py", line 864, in process_wsi
    keep_idx = self._post_process_edge_cells(cell_list=cell_dict_wsi)
  File "/nemo/lab/swantonc/working/liul/.conda/envs/cellvit_env/lib/python3.10/site-packages/cellvit/inference/inference.py", line 507, in _post_process_edge_cells
    cleaned_cells = cell_cleaner.clean_detected_cells()
  File "/nemo/lab/swantonc/working/liul/.conda/envs/cellvit_env/lib/python3.10/site-packages/cellvit/inference/overlap_cell_cleaner.py", line 76, in clean_detected_cells
    cleaned_edge_cells = self._remove_overlap(cleaned_edge_cells)
  File "/nemo/lab/swantonc/working/liul/.conda/envs/cellvit_env/lib/python3.10/site-packages/cellvit/inference/overlap_cell_cleaner.py", line 141, in _remove_overlap
    merged_cells = self._remove_overlap_shapely_1_8(
  File "/nemo/lab/swantonc/working/liul/.conda/envs/cellvit_env/lib/python3.10/site-packages/cellvit/inference/overlap_cell_cleaner.py", line 191, in _remove_overlap_shapely_1_8
    if query_poly.uid not in iterated_cells:
AttributeError: 'Polygon' object has no attribute 'uid'

This is my version of Shapely confirmed:

$pip show shapely
Name: Shapely
Version: 1.8.5.post1

This was the command run:

cellvit-inference \
    --model SAM \
    --nuclei_taxonomy pannuke \
    --batch_size 8 \
    --cpu_count 32 \
    --ray_worker 6 \
    --ray_remote_cpus 4 \
    --outdir ./test/JP2K-33003-2/ \
    process_wsi \
    --wsi_path ./test/JP2K-33003-2/JP2K-33003-2.svs \
    --wsi_mpp 0.25 \
    --wsi_magnification 40

At a glance, this seems to be because the uid attribute has been commented out from the function?

    def _remove_overlap_shapely_1_8(
        self, cleaned_edge_cells: pd.DataFrame
    ) -> pd.DataFrame:
        """Remove overlapping cells from provided DataFrame

        Args:
            cleaned_edge_cells (pd.DataFrame): DataFrame that should be cleaned

        Returns:
            pd.DataFrame: Cleaned DataFrame
        """
        merged_cells = cleaned_edge_cells

        for iteration in range(20):
            poly_list = []
            for idx, cell_info in merged_cells.iterrows():
                poly = Polygon(cell_info["contour"])
                if not poly.is_valid:
                    self.logger.debug("Found invalid polygon - Fixing with buffer 0")
                    multi = poly.buffer(0)
                    if isinstance(multi, MultiPolygon):
                        if len(multi) > 1:
                            poly_idx = np.argmax([p.area for p in multi])
                            poly = multi[poly_idx]
                            poly = Polygon(poly)
                        else:
                            poly = multi[0]
                            poly = Polygon(poly)
                    else:
                        poly = Polygon(multi)
                # poly.uid = idx
                poly_list.append(poly)

            # use an strtree for fast querying
            tree = strtree.STRtree(poly_list)

            merged_idx = deque()
            iterated_cells = set()
            overlaps = 0

            for query_poly in poly_list:
                if query_poly.uid not in iterated_cells:

Uncommenting # poly.uid = idx solved the issue

Does the developer recommend upgrading to Shapely version >= 2.0.0?

[FabianHoerst]

We need more time to investigate this, as we are unable to reproduce the error with our files. It is worth a try to use shapely >=2.0.0, as i think the uid has been removed later on.