Refactor v0.1

README.md

@@ -0,0 +1,50 @@
+## Still in development
+
+#  cellmap-flow 
+
+```bash
+$ cellmap_flow_server
+
+Usage: cellmap_flow_server [OPTIONS] COMMAND [ARGS]...
+
+  Examples:     
+    To use Dacapo run the following commands:  
+    cellmap_flow_server dacapo -r my_run -i iteration -d data_path
+
+    To use custom script
+    cellmap_flow_server script -s script_path -d data_path
+
+    To use bioimage-io model 
+    cellmap_flow_server bioimage -m model_path -d data_path
+
+
+Commands:
+  bioimage  Run the CellMapFlow server with a bioimage-io model.
+  dacapo    Run the CellMapFlow server with a DaCapo model.
+  script    Run the CellMapFlow server with a custom script.
+```
+
+
+Currently available:
+## Using custom script:
+which enable using any model by providing a script e.g. [example/model_spec.py](example/model_spec.py)
+e.g.
+```bash
+cellmap_flow script -c /groups/cellmap/cellmap/zouinkhim/cellmap-flow/example/model_spec.py -d /nrs/cellmap/data/jrc_mus-cerebellum-1/jrc_mus-cerebellum-1.zarr/recon-1/em/fibsem-uint8/s0 
+```
+
+## Using Dacapo model:
+which enable inference using a Dacapo model by providing the run name and iteration number
+e.g.
+```bash
+cellmap_flow dacapo -r 20241204_finetune_mito_affs_task_datasplit_v3_u21_kidney_mito_default_cache_8_1 -i 700000 -d /nrs/cellmap/data/jrc_ut21-1413-003/jrc_ut21-1413-003.zarr/recon-1/em/fibsem-uint8/s0"
+```
+
+## Using bioimage-io model:
+still in development
+
+
+## Limitation:
+Currently only supporting data locating in /nrs/cellmap or /groups/cellmap because there is a data server already implemented for them.
+
+

cellmap_flow/__init__.py

@@ -0,0 +1,2 @@
+__version__ = "0.1.0"
+__version_info__ = tuple(int(i) for i in __version__.split("."))

cellmap_flow/bioimagezoo_processor.py

@@ -0,0 +1,69 @@
+# %%
+from bioimageio.core import load_description
+from bioimageio.core import predict  # , predict_many
+
+# %%
+from bioimageio.core import Tensor
+import numpy as np
+from bioimageio.core import Sample
+from bioimageio.core.digest_spec import get_member_ids
+
+
+# create random input tensor
+def process_chunk(model, idi, roi):
+    input_image = idi.to_ndarray_ts(roi)
+    # add dimensions to start of input image
+    # print(input_image.shape)
+    # input_image = block_reduce(input_image, (4,1,1), np.mean, cval=0)
+    # print(input_image.shape)
+    # input_image = (input_image - np.min(input_image))/np.max(input_image)
+    # input_image = np.random.rand(1, 1, 64, 64, 64).astype(np.float32)
+    if len(model.outputs[0].axes) == 5:
+        input_image = input_image[np.newaxis, np.newaxis, ...].astype(np.float32)
+        test_input_tensor = Tensor.from_numpy(
+            input_image, dims=["batch", "c", "z", "y", "x"]
+        )
+    else:
+        input_image = input_image[:, np.newaxis, ...].astype(np.float32)
+
+        test_input_tensor = Tensor.from_numpy(
+            input_image, dims=["batch", "c", "y", "x"]
+        )
+    sample_input_id = get_member_ids(model.inputs)[0]
+    sample_output_id = get_member_ids(model.outputs)[0]
+
+    sample = Sample(
+        members={sample_input_id: test_input_tensor}, stat={}, id="sample-from-numpy"
+    )
+    prediction: Sample = predict(
+        model=model, inputs=sample, skip_preprocessing=sample.stat is not None
+    )
+    ndim = prediction.members[sample_output_id].data.ndim
+    output = prediction.members[sample_output_id].data.to_numpy()
+    if ndim < 5 and len(model.outputs) > 1:
+        if len(model.outputs) > 1:
+            outputs = []
+            for id in get_member_ids(model.outputs):
+                output = prediction.members[id].data.to_numpy()
+                if output.ndim == 3:
+                    output = output[:, np.newaxis, ...]
+                outputs.append(output)
+            output = np.concatenate(outputs, axis=1)
+        output = np.ascontiguousarray(np.swapaxes(output, 1, 0))
+
+    else:
+        output = output[0, ...]
+    # if ndim == 5:
+    #     # then is b,c,z,y,x, and only want z,y,x
+    #     output = 255 * output[0, ...]
+    # elif ndim == 4:
+    #     # then is b,c,y,x  since it is 2d which is really z,c,y,x
+    #     output = 255 * output[0, 0, ...]
+    output = 255 * output
+    output = output.astype(np.uint8)
+    return output
+
+
+# %%
+# import torch
+# from pydeepimagej.yaml_utils import create_model_yaml