If annotations don't have enough slices...

[azatian]

Dear ScalableMinds,

I have been attempting to download my annotations as a segmentation mask from our server's WK instance. And I believe I have run into an issue that I can't solve. I think it's related to the underlying zarr file and how the chunk size is set up.

import webknossos as wk
dataset = wk.Dataset.open("our-zarr-link")
MAG = wk.Mag("1")
SEGMENT_IDS = [1]
mag_view = dataset.get_segmentation_layers()[0].get_mag(MAG)
the_zarr = mag_view.get_zarr_array()
the_zarr.info()

This is the output of the_zarr.info()

Type zarr.core.Array Data type uint32 Shape (1, 912, 912, 6) Chunk shape (1, 32, 32, 32) Order F Read-only False Compressor None Store type zarr.storage.FSStore No. bytes 19961856 (19.0M) No. bytes stored 161 Storage ratio 123986.7 Chunks initialized 0/841

Every time I try to do anything with mag_view including the functions read and get_buffered_slice_reader as well as any zarr array operations, even rechunking, I get the following error:

ValueError: cannot reshape array of size 0 into shape (1,32,32,32)

I tried to create a new mag with a smaller chunk size, one that fits with my data, and it seems like I can't add a mag on a layer that is hosted as a zarr link. I get this error.

ClientResponseError: 400, message='Bad Request', url=URL('our-zarr-link/Volume/2/.zarray')

pip show webknossos Version: 0.10.24

[hotzenklotz]

Thanks for the bug report. Have you tried updating to the latest version of the webknossos lib v0.12.3? Your package is almost a year old.

I just tried your example from above and I have no trouble with mag_view.read() with the latest version, i.e.:

import webknossos as wk

dataset = wk.Dataset.open_remote(
        dataset_name_or_url="l4dense_motta_et_al_demo_v2",
        organization_id="scalable_minds",
    )

MAG = wk.Mag("1")
SEGMENT_IDS = [1]
mag_view = dataset.get_segmentation_layers()[0].get_mag(MAG)
the_zarr = mag_view.get_zarr_array()
the_zarr.info

data = mag_view.read(absolute_bounding_box=wk.BoundingBox((10, 10, 10), (10, 10,10))) 
data.shape

>>> (1, 10, 10, 10)

[azatian]

Hi, thank you so much for your quick reply! I just tried with the latest version of webknossos on Python and I'm still getting the same issues. Please see below:

import webknossos as wk
import pandas as pd
import numpy as np
# posting our link for complete visibility, but will need to remove later
dataset = wk.Dataset.open("http://catmaid2.hms.harvard.edu:9000/data/annotations/zarr/4pJXHDyVK_rvi7db") 
MAG = wk.Mag("1")
SEGMENT_IDS = [1]
mag_view = dataset2.get_segmentation_layers()[0].get_mag(MAG)
the_zarr = mag_view.get_zarr_array()
the_zarr
>>> <zarr.core.Array (1, 912, 912, 6) uint32>
the_zarr.shape
>>> (1, 912, 912, 6)
the_zarr[0,:,:,:]
>>> See Error 1
data = mag_view.read() 
>>> See Error 2
data = mag_view.read(absolute_bounding_box=wk.BoundingBox((0,0,0), (10, 10,1))) 
>>> See Error 3

Error 1 ValueError Traceback (most recent call last) Input In [10], in <cell line: 1>() ----> 1 the_zarr[0,:,:,:]

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:821, in Array.getitem(self, selection) 819 result = self.vindex[selection] 820 else: --> 821 result = self.get_basic_selection(pure_selection, fields=fields) 822 return result

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:947, in Array.get_basic_selection(self, selection, out, fields) 944 return self._get_basic_selection_zd(selection=selection, out=out, 945 fields=fields) 946 else: --> 947 return self._get_basic_selection_nd(selection=selection, out=out, 948 fields=fields)

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:990, in Array._get_basic_selection_nd(self, selection, out, fields) 984 def _get_basic_selection_nd(self, selection, out=None, fields=None): 985 # implementation of basic selection for array with at least one dimension 986 987 # setup indexer 988 indexer = BasicIndexer(selection, self) --> 990 return self._get_selection(indexer=indexer, out=out, fields=fields)

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:1290, in Array._get_selection(self, indexer, out, fields) 1287 else: 1288 # allow storage to get multiple items at once 1289 lchunk_coords, lchunk_selection, lout_selection = zip(*indexer) -> 1290 self._chunk_getitems(lchunk_coords, lchunk_selection, out, lout_selection, 1291 drop_axes=indexer.drop_axes, fields=fields) 1293 if out.shape: 1294 return out

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:2063, in Array._chunk_getitems(self, lchunk_coords, lchunk_selection, out, lout_selection, drop_axes, fields) 2061 for ckey, chunk_select, out_select in zip(ckeys, lchunk_selection, lout_selection): 2062 if ckey in cdatas: -> 2063 self._process_chunk( 2064 out, 2065 cdatas[ckey], 2066 chunk_select, 2067 drop_axes, 2068 out_is_ndarray, 2069 fields, 2070 out_select, 2071 partial_read_decode=partial_read_decode, 2072 ) 2073 else: 2074 # check exception type 2075 if self._fill_value is not None:

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:1949, in Array._process_chunk(self, out, cdata, chunk_selection, drop_axes, out_is_ndarray, fields, out_selection, partial_read_decode) 1947 except ArrayIndexError: 1948 cdata = cdata.read_full() -> 1949 chunk = self._decode_chunk(cdata) 1951 # select data from chunk 1952 if fields:

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:2258, in Array._decode_chunk(self, cdata, start, nitems, expected_shape) 2256 # ensure correct chunk shape 2257 chunk = chunk.reshape(-1, order='A') -> 2258 chunk = chunk.reshape(expected_shape or self._chunks, order=self._order) 2260 return chunk

ValueError: cannot reshape array of size 0 into shape (1,32,32,32)

Error 2

---

ValueError Traceback (most recent call last) Input In [11], in <cell line: 1>() ----> 1 data = mag_view.read()

File ~/Desktop/code/base/lib/python3.9/site-packages/webknossos/dataset/mag_view.py:209, in MagView.read(self, offset, size, relative_offset, absolute_offset, relative_bounding_box, absolute_bounding_box) 182 def read( 183 self, 184 offset: Optional[Vec3IntLike] = None, # deprecated, relative, in current mag (...) 193 ) -> np.ndarray: 194 # THIS METHOD CAN BE REMOVED WHEN THE DEPRECATED OFFSET IS REMOVED 196 if ( 197 relative_offset is not None 198 or absolute_offset is not None (...) 207 and relative_bounding_box is None 208 ): --> 209 return super().read( 210 offset, 211 size, 212 relative_offset=relative_offset, 213 absolute_offset=absolute_offset, 214 absolute_bounding_box=absolute_bounding_box, 215 relative_bounding_box=relative_bounding_box, 216 ) 217 else: 218 with warnings.catch_warnings():

File ~/Desktop/code/base/lib/python3.9/site-packages/webknossos/dataset/view.py:403, in View.read(self, offset, size, relative_offset, absolute_offset, relative_bounding_box, absolute_bounding_box) 395 assert not mag1_bbox.is_empty(), ( 396 f"The size ({mag1_bbox.size} in mag1) contains a zero. " 397 + "All dimensions must be strictly larger than '0'." 398 ) 399 assert ( 400 mag1_bbox.topleft.is_positive() 401 ), f"The offset ({mag1_bbox.topleft} in mag1) must not contain negative dimensions." --> 403 return self._read_without_checks(mag1_bbox.in_mag(self._mag))

File ~/Desktop/code/base/lib/python3.9/site-packages/webknossos/dataset/view.py:440, in View._read_without_checks(self, current_mag_bbox) 436 def _read_without_checks( 437 self, 438 current_mag_bbox: BoundingBox, 439 ) -> np.ndarray: --> 440 data = self._array.read( 441 current_mag_bbox.topleft.to_np(), current_mag_bbox.size.to_np() 442 ) 443 return data

File ~/Desktop/code/base/lib/python3.9/site-packages/webknossos/dataset/_array.py:343, in ZarrArray.read(self, offset, shape) 341 zarray = self._zarray 342 with _blosc_disable_threading(): --> 343 data = zarray[ 344 :, 345 offset.x : (offset.x + shape.x), 346 offset.y : (offset.y + shape.y), 347 offset.z : (offset.z + shape.z), 348 ] 349 if data.shape != shape: 350 padded_data = np.zeros( 351 (self.info.num_channels,) + shape.to_tuple(), dtype=data.dtype 352 )

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:821, in Array.getitem(self, selection) 819 result = self.vindex[selection] 820 else: --> 821 result = self.get_basic_selection(pure_selection, fields=fields) 822 return result

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:947, in Array.get_basic_selection(self, selection, out, fields) 944 return self._get_basic_selection_zd(selection=selection, out=out, 945 fields=fields) 946 else: --> 947 return self._get_basic_selection_nd(selection=selection, out=out, 948 fields=fields)

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:990, in Array._get_basic_selection_nd(self, selection, out, fields) 984 def _get_basic_selection_nd(self, selection, out=None, fields=None): 985 # implementation of basic selection for array with at least one dimension 986 987 # setup indexer 988 indexer = BasicIndexer(selection, self) --> 990 return self._get_selection(indexer=indexer, out=out, fields=fields)

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:1290, in Array._get_selection(self, indexer, out, fields) 1287 else: 1288 # allow storage to get multiple items at once 1289 lchunk_coords, lchunk_selection, lout_selection = zip(*indexer) -> 1290 self._chunk_getitems(lchunk_coords, lchunk_selection, out, lout_selection, 1291 drop_axes=indexer.drop_axes, fields=fields) 1293 if out.shape: 1294 return out

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:2063, in Array._chunk_getitems(self, lchunk_coords, lchunk_selection, out, lout_selection, drop_axes, fields) 2061 for ckey, chunk_select, out_select in zip(ckeys, lchunk_selection, lout_selection): 2062 if ckey in cdatas: -> 2063 self._process_chunk( 2064 out, 2065 cdatas[ckey], 2066 chunk_select, 2067 drop_axes, 2068 out_is_ndarray, 2069 fields, 2070 out_select, 2071 partial_read_decode=partial_read_decode, 2072 ) 2073 else: 2074 # check exception type 2075 if self._fill_value is not None:

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:1949, in Array._process_chunk(self, out, cdata, chunk_selection, drop_axes, out_is_ndarray, fields, out_selection, partial_read_decode) 1947 except ArrayIndexError: 1948 cdata = cdata.read_full() -> 1949 chunk = self._decode_chunk(cdata) 1951 # select data from chunk 1952 if fields:

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:2258, in Array._decode_chunk(self, cdata, start, nitems, expected_shape) 2256 # ensure correct chunk shape 2257 chunk = chunk.reshape(-1, order='A') -> 2258 chunk = chunk.reshape(expected_shape or self._chunks, order=self._order) 2260 return chunk

ValueError: cannot reshape array of size 0 into shape (1,32,32,32)

Error 3

---

ValueError Traceback (most recent call last) Input In [12], in <cell line: 1>() ----> 1 data = mag_view.read(absolute_bounding_box=wk.BoundingBox((0,0,0), (10, 10,1)))

File ~/Desktop/code/base/lib/python3.9/site-packages/webknossos/dataset/mag_view.py:209, in MagView.read(self, offset, size, relative_offset, absolute_offset, relative_bounding_box, absolute_bounding_box) 182 def read( 183 self, 184 offset: Optional[Vec3IntLike] = None, # deprecated, relative, in current mag (...) 193 ) -> np.ndarray: 194 # THIS METHOD CAN BE REMOVED WHEN THE DEPRECATED OFFSET IS REMOVED 196 if ( 197 relative_offset is not None 198 or absolute_offset is not None (...) 207 and relative_bounding_box is None 208 ): --> 209 return super().read( 210 offset, 211 size, 212 relative_offset=relative_offset, 213 absolute_offset=absolute_offset, 214 absolute_bounding_box=absolute_bounding_box, 215 relative_bounding_box=relative_bounding_box, 216 ) 217 else: 218 with warnings.catch_warnings():

File ~/Desktop/code/base/lib/python3.9/site-packages/webknossos/dataset/view.py:403, in View.read(self, offset, size, relative_offset, absolute_offset, relative_bounding_box, absolute_bounding_box) 395 assert not mag1_bbox.is_empty(), ( 396 f"The size ({mag1_bbox.size} in mag1) contains a zero. " 397 + "All dimensions must be strictly larger than '0'." 398 ) 399 assert ( 400 mag1_bbox.topleft.is_positive() 401 ), f"The offset ({mag1_bbox.topleft} in mag1) must not contain negative dimensions." --> 403 return self._read_without_checks(mag1_bbox.in_mag(self._mag))

File ~/Desktop/code/base/lib/python3.9/site-packages/webknossos/dataset/view.py:440, in View._read_without_checks(self, current_mag_bbox) 436 def _read_without_checks( 437 self, 438 current_mag_bbox: BoundingBox, 439 ) -> np.ndarray: --> 440 data = self._array.read( 441 current_mag_bbox.topleft.to_np(), current_mag_bbox.size.to_np() 442 ) 443 return data

File ~/Desktop/code/base/lib/python3.9/site-packages/webknossos/dataset/_array.py:343, in ZarrArray.read(self, offset, shape) 341 zarray = self._zarray 342 with _blosc_disable_threading(): --> 343 data = zarray[ 344 :, 345 offset.x : (offset.x + shape.x), 346 offset.y : (offset.y + shape.y), 347 offset.z : (offset.z + shape.z), 348 ] 349 if data.shape != shape: 350 padded_data = np.zeros( 351 (self.info.num_channels,) + shape.to_tuple(), dtype=data.dtype 352 )

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:821, in Array.getitem(self, selection) 819 result = self.vindex[selection] 820 else: --> 821 result = self.get_basic_selection(pure_selection, fields=fields) 822 return result

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:947, in Array.get_basic_selection(self, selection, out, fields) 944 return self._get_basic_selection_zd(selection=selection, out=out, 945 fields=fields) 946 else: --> 947 return self._get_basic_selection_nd(selection=selection, out=out, 948 fields=fields)

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:990, in Array._get_basic_selection_nd(self, selection, out, fields) 984 def _get_basic_selection_nd(self, selection, out=None, fields=None): 985 # implementation of basic selection for array with at least one dimension 986 987 # setup indexer 988 indexer = BasicIndexer(selection, self) --> 990 return self._get_selection(indexer=indexer, out=out, fields=fields)

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:1290, in Array._get_selection(self, indexer, out, fields) 1287 else: 1288 # allow storage to get multiple items at once 1289 lchunk_coords, lchunk_selection, lout_selection = zip(*indexer) -> 1290 self._chunk_getitems(lchunk_coords, lchunk_selection, out, lout_selection, 1291 drop_axes=indexer.drop_axes, fields=fields) 1293 if out.shape: 1294 return out

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:2063, in Array._chunk_getitems(self, lchunk_coords, lchunk_selection, out, lout_selection, drop_axes, fields) 2061 for ckey, chunk_select, out_select in zip(ckeys, lchunk_selection, lout_selection): 2062 if ckey in cdatas: -> 2063 self._process_chunk( 2064 out, 2065 cdatas[ckey], 2066 chunk_select, 2067 drop_axes, 2068 out_is_ndarray, 2069 fields, 2070 out_select, 2071 partial_read_decode=partial_read_decode, 2072 ) 2073 else: 2074 # check exception type 2075 if self._fill_value is not None:

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:1949, in Array._process_chunk(self, out, cdata, chunk_selection, drop_axes, out_is_ndarray, fields, out_selection, partial_read_decode) 1947 except ArrayIndexError: 1948 cdata = cdata.read_full() -> 1949 chunk = self._decode_chunk(cdata) 1951 # select data from chunk 1952 if fields:

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:2258, in Array._decode_chunk(self, cdata, start, nitems, expected_shape) 2256 # ensure correct chunk shape 2257 chunk = chunk.reshape(-1, order='A') -> 2258 chunk = chunk.reshape(expected_shape or self._chunks, order=self._order) 2260 return chunk

ValueError: cannot reshape array of size 0 into shape (1,32,32,32)

​

[hotzenklotz]

Mhm, I believe the error stems from the line:

dataset = wk.Dataset.open("http://catmaid2.hms.harvard.edu:9000/data/annotations/zarr/4pJXHDyVK_rvi7db") 

The webknossos-libs are meant to interface with WEBKNOSSOS URI for datasets and annotations, i.e. https://<wkserver>.com/annotations/6124d30a010000ab009167ed or https://<wkserver>.com/datasets/scalable_minds/110629_k0725_segmentation_v1/view.

The libs do not support loading/streaming Zarr links. Please directly use a Zarr library such as zarr-python for that. Try

import zarr

color_layer = zarr.open_group("http://catmaid2.hms.harvard.edu:9000/data/annotations/zarr/4pJXHDyVK_rvi7db")
color_layer.tree()

>> / <zarr.hierarchy.Group '/shape'> <zarr.hierarchy.Group '/dtype'>
 ├── .zgroup
 ├── Volume
 │   └── 1 (1, 912, 912, 6) uint32
 ├── cellseg
 │   └── 1 (1, 912, 912, 6) uint8
 ├── datasource-properties.json
 └── img
     └── 1 (1, 912, 912, 6) uint8

color_layer["/img/1"].shape
>> (1, 912, 912, 6)

For compatibility and interoperability purposes with other tools WK can expose it's data as Zarr-API compatbile URIs through the UI (as you did) or by exposing the raw zarr-object through the libs for working zarr-python or dask.

[azatian]

Thank you! I tried to use thezarr-python package....and I ran into the same issue when trying to view the annotation layer, which is Volume. The other layers work perfectly. Please see below.

color_layer["/Volume/1"].shape
>>> (1, 912, 912, 6)
color_layer["/Volume/1"][0,:,:,0]
>>> See Error 1

Error 1

---

ValueError Traceback (most recent call last) Input In [17], in <cell line: 1>() ----> 1 color_layer["/Volume/1"][0,:,:,0]

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:821, in Array.getitem(self, selection) 819 result = self.vindex[selection] 820 else: --> 821 result = self.get_basic_selection(pure_selection, fields=fields) 822 return result

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:947, in Array.get_basic_selection(self, selection, out, fields) 944 return self._get_basic_selection_zd(selection=selection, out=out, 945 fields=fields) 946 else: --> 947 return self._get_basic_selection_nd(selection=selection, out=out, 948 fields=fields)

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:990, in Array._get_basic_selection_nd(self, selection, out, fields) 984 def _get_basic_selection_nd(self, selection, out=None, fields=None): 985 # implementation of basic selection for array with at least one dimension 986 987 # setup indexer 988 indexer = BasicIndexer(selection, self) --> 990 return self._get_selection(indexer=indexer, out=out, fields=fields)

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:1290, in Array._get_selection(self, indexer, out, fields) 1287 else: 1288 # allow storage to get multiple items at once 1289 lchunk_coords, lchunk_selection, lout_selection = zip(*indexer) -> 1290 self._chunk_getitems(lchunk_coords, lchunk_selection, out, lout_selection, 1291 drop_axes=indexer.drop_axes, fields=fields) 1293 if out.shape: 1294 return out

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:2063, in Array._chunk_getitems(self, lchunk_coords, lchunk_selection, out, lout_selection, drop_axes, fields) 2061 for ckey, chunk_select, out_select in zip(ckeys, lchunk_selection, lout_selection): 2062 if ckey in cdatas: -> 2063 self._process_chunk( 2064 out, 2065 cdatas[ckey], 2066 chunk_select, 2067 drop_axes, 2068 out_is_ndarray, 2069 fields, 2070 out_select, 2071 partial_read_decode=partial_read_decode, 2072 ) 2073 else: 2074 # check exception type 2075 if self._fill_value is not None:

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:1949, in Array._process_chunk(self, out, cdata, chunk_selection, drop_axes, out_is_ndarray, fields, out_selection, partial_read_decode) 1947 except ArrayIndexError: 1948 cdata = cdata.read_full() -> 1949 chunk = self._decode_chunk(cdata) 1951 # select data from chunk 1952 if fields:

File ~/Desktop/code/base/lib/python3.9/site-packages/zarr/core.py:2258, in Array._decode_chunk(self, cdata, start, nitems, expected_shape) 2256 # ensure correct chunk shape 2257 chunk = chunk.reshape(-1, order='A') -> 2258 chunk = chunk.reshape(expected_shape or self._chunks, order=self._order) 2260 return chunk

ValueError: cannot reshape array of size 0 into shape (1,32,32,32)

[hotzenklotz]

Thanks for the detailed listing. For completeness sake, what version or release of WEBKNOSSOS are you running?

You can find the version number in the "help" section of the navbar:

[azatian]

Hello! We are using version 22.09.0 We currently have issues with the Python client finding our annotations as datasets, so that is why I decided to try the zarr link route. I can post about that issue as well, but didn't want to convolute too many issues together.

[hotzenklotz]

We are using version 22.09.0

That version is more than half a year old. In this timeframe we have worked extensively with the Zarr community and improved the stability and compatibility of the Zarr APIs in WK. While I can not promise that an update will resolve your issue I am fairly confident that it might.

We currently have issues with the Python client finding our annotations as datasets

Interesting. Let's hear it. (Feel free to open a second issues/thread for that). Generally speaking Annotations ≠ Datasets. A dataset typically contains the raw microscopy data, metadata, access permissions etc. An annotation typically contains a reference to the dataset and any new volume and skeleton labels/makers that you placed. A dataset can usually have several annotations "connected" to it, e.g. 10 users doing one annotation for the same dataset (1:n relationship). Unfortunately, in reality the border between the two concepts is sometime very "fluid" and perhaps the naming of our APIs needs to be clearer. Feedback is always welcome :-)

Hence, there are different webknossos-libs Python object for Annotations and Datasets. If you only interested in the volume annotation of the dataset (and assuming that the volume annotations were created from scratch on a "blank" surface), I recommend the follow code examples:

• https://docs.webknossos.org/webknossos-py/examples/download_segments.html
• https://docs.webknossos.org/webknossos-py/examples/learned_segmenter.html