Merge pull request #181 from bioimage-io/stuff

Small post-release clean up

Author: constantinpape

README.md

@@ -1,12 +1,17 @@
 # core-bioimage-io-python
 
-Python specific core utilities for running models in the [BioImage Model Zoo](https://bioimage.io)
+Python specific core utilities for running models in the [BioImage Model Zoo](https://bioimage.io).
 
 ## Installation
 
 ### Via Conda
 
-The `bioimageio.core` package supports various back-ends for running BioimageIO networks:
+The `bioimageio.core` package can be installed from conda-forge via
+```
+conda install -c conda-forge bioimageio.core
+```
+if you don't install any additional deep learning libraries, you will only be able to use general convenience functionality, but not any functionality for model prediction.
+To install additional deep learning libraries use:
 
 * Pytorch/Torchscript:
   ```bash
@@ -69,6 +74,13 @@ Run prediction for an image stored on disc:
 bioimageio predict-image -m <MODEL> -i <INPUT> -o <OUTPUT>
 ```
 
+Run prediction for multiple images stored on disc:
+```
+bioimagei predict-images -m <MODEL> -i <INPUT_PATTERN> - o <OUTPUT_FOLDER>
+```
+`<INPUT_PATTERN>` is a `glob` pattern to select the desired images, e.g. `/path/to/my/images/*.tif`.
+
+
 ## From python
 
 `bioimageio.core` can be used as a python library. See the notebook [example/bioimageio-core-usage.ipynb](https://github.com/bioimage-io/core-bioimage-io-python/blob/main/example/bioimageio-core-usage.ipynb) for usage examples.

bioimageio/core/weight_converter/keras/tensorflow.py

@@ -13,13 +13,13 @@
 
 # adapted from
 # https://github.com/deepimagej/pydeepimagej/blob/master/pydeepimagej/yaml/create_config.py#L236
-def _convert_tf1(keras_weight_path, output_path, zip_weights):
+def _convert_tf1(keras_weight_path, output_path, input_name, output_name, zip_weights):
     def build_tf_model():
         keras_model = keras.models.load_model(keras_weight_path)
 
         builder = saved_model.builder.SavedModelBuilder(output_path)
         signature = saved_model.signature_def_utils.predict_signature_def(
-            inputs={"input": keras_model.input}, outputs={"output": keras_model.output}
+            inputs={input_name: keras_model.input}, outputs={output_name: keras_model.output}
         )
 
         signature_def_map = {saved_model.signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY: signature}
@@ -98,6 +98,10 @@ def convert_weights_to_tensorflow_saved_model_bundle(
             raise RuntimeError(f"Tensorflow major versions of model {model_tf_major_ver} is not {tf_major_ver}")
 
     if tf_major_ver == 1:
-        return _convert_tf1(weight_path, str(path_), zip_weights)
+        if len(model.inputs) != 1 or len(model.outputs) != 1:
+            raise NotImplementedError(
+                "Weight conversion for models with multiple inputs or outputs is not yet implemented."
+            )
+        return _convert_tf1(weight_path, str(path_), model.inputs[0].name, model.outputs[0].name, zip_weights)
     else:
-        raise NotImplementedError
+        raise NotImplementedError("Weight conversion for tensorflow 2 is not yet implemented.")

tests/build_spec/test_build_spec.py

@@ -1,3 +1,4 @@
+import os
 import bioimageio.spec as spec
 from bioimageio.core import load_raw_resource_description, load_resource_description
 from bioimageio.core.resource_io import nodes
@@ -14,6 +15,7 @@ def _test_build_spec(
     use_implicit_output_shape=False,
     add_deepimagej_config=False,
     use_original_covers=False,
+    use_absoloute_arch_path=False,
 ):
     from bioimageio.core.build_spec import build_model
 
@@ -28,6 +30,11 @@ def _test_build_spec(
         weight_spec = model_spec.weights["pytorch_state_dict"]
         model_kwargs = None if weight_spec.kwargs is missing else weight_spec.kwargs
         architecture = str(weight_spec.architecture)
+        if use_absoloute_arch_path:
+            arch_path, cls_name = architecture.split(":")
+            arch_path = os.path.abspath(os.path.join(root, arch_path))
+            assert os.path.exists(arch_path)
+            architecture = f"{arch_path}:{cls_name}"
         weight_type_ = None  # the weight type can be auto-detected
     elif weight_type == "torchscript":
         architecture = None
@@ -152,3 +159,10 @@ def test_build_spec_deepimagej_keras(unet2d_keras, tmp_path):
 # test with original covers
 def test_build_spec_with_original_covers(unet2d_nuclei_broad_model, tmp_path):
     _test_build_spec(unet2d_nuclei_broad_model, tmp_path / "model.zip", "torchscript", use_original_covers=True)
+
+
+# test with absolute path for the architecture file
+def test_build_spec_abs_arch_path(unet2d_nuclei_broad_model, tmp_path):
+    _test_build_spec(
+        unet2d_nuclei_broad_model, tmp_path / "model.zip", "pytorch_state_dict", use_absoloute_arch_path=True
+    )