Mercurial > repos > bgruening > bioimage_inference
changeset 0:caea9ee1ffac draft
planemo upload for repository https://github.com/bgruening/galaxytools/tree/recommendation_training/tools/bioimaging commit 57f46739f4365f59cd52c515bdd3fae2e01b734e
| author | bgruening |
|---|---|
| date | Fri, 02 Aug 2024 15:40:35 +0000 |
| parents | |
| children | b0f2d3b4a278 |
| files | bioimage_inference.xml main.py test-data/input_nucleisegboundarymodel.png test-data/output_nucleisegboundarymodel.tif test-data/output_nucleisegboundarymodel_matrix.npy |
| diffstat | 5 files changed, 166 insertions(+), 0 deletions(-) [+] |
line wrap: on
line diff
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/bioimage_inference.xml Fri Aug 02 15:40:35 2024 +0000 @@ -0,0 +1,80 @@ +<tool id="bioimage_inference" name="Process image using a BioImage.IO model" version="@TOOL_VERSION@+galaxy@VERSION_SUFFIX@" profile="23.0"> + <description>with PyTorch</description> + <macros> + <token name="@TOOL_VERSION@">2.3.1</token> + <token name="@VERSION_SUFFIX@">0</token> + </macros> + <creator> + <organization name="European Galaxy Team" url="https://galaxyproject.org/eu/" /> + <person givenName="Anup" familyName="Kumar" email="kumara@informatik.uni-freiburg.de" /> + <person givenName="Beatriz" familyName="Serrano-Solano" email="beatriz.serrano.solano@eurobioimaging.eu" /> + <person givenName="Leonid" familyName="Kostrykin" email="leonid.kostrykin@bioquant.uni-heidelberg.de" /> + </creator> + <edam_operations> + <edam_operation>operation_3443</edam_operation> + </edam_operations> + <xrefs> + <xref type="bio.tools">pytorch</xref> + <xref type="biii">pytorch</xref> + </xrefs> + <requirements> + <requirement type="package" version="3.9.12">python</requirement> + <requirement type="package" version="@TOOL_VERSION@">pytorch</requirement> + <requirement type="package" version="0.18.1">torchvision</requirement> + <requirement type="package" version="2.34.2">imageio</requirement> + </requirements> + <version_command>echo "@VERSION@"</version_command> + <command detect_errors="aggressive"> + <![CDATA[ + python '$__tool_directory__/main.py' + --imaging_model '$input_imaging_model' + --image_file '$input_image_file' + --image_size '$input_image_input_size' + ]]> + </command> + <inputs> + <param name="input_imaging_model" type="data" format="zip" label="BioImage.IO model" help="Please upload a BioImage.IO model."/> + <param name="input_image_file" type="data" format="tiff,png" label="Input image" help="Please provide an input image for the analysis."/> + <param name="input_image_input_size" type="text" label="Size of the input image" help="Provide the size of the input image. See the chosen model's RDF file to find the correct input size. For example: for the BioImage.IO model MitochondriaEMSegmentationBoundaryModel, the input size is 256 x 256 x 32 x 1. Enter the size as 256,256,32,1."/> + </inputs> + <outputs> + <data format="tif" name="output_predicted_image" from_work_dir="output_predicted_image.tif" label="Predicted image"></data> + <data format="npy" name="output_predicted_image_matrix" from_work_dir="output_predicted_image_matrix.npy" label="Predicted image tensor"></data> + </outputs> + <tests> + <test> + <param name="input_imaging_model" value="input_imaging_model.zip" location="https://zenodo.org/api/records/6647674/files/weights-torchscript.pt/content"/> + <param name="input_image_file" value="input_image_file.tif" location="https://zenodo.org/api/records/6647674/files/sample_input_0.tif/content"/> + <param name="input_image_input_size" value="256,256,1,1"/> + <output name="output_predicted_image" file="output_nucleisegboundarymodel.tif" compare="sim_size" delta="100" /> + <output name="output_predicted_image_matrix" file="output_nucleisegboundarymodel_matrix.npy" compare="sim_size" delta="100" /> + </test> + <test> + <param name="input_imaging_model" value="input_imaging_model.zip" location="https://zenodo.org/api/records/6647674/files/weights-torchscript.pt/content"/> + <param name="input_image_file" value="input_nucleisegboundarymodel.png"/> + <param name="input_image_input_size" value="256,256,1,1"/> + <output name="output_predicted_image" file="output_nucleisegboundarymodel.tif" compare="sim_size" delta="100" /> + <output name="output_predicted_image_matrix" file="output_nucleisegboundarymodel_matrix.npy" compare="sim_size" delta="100" /> + </test> + </tests> + <help> + <![CDATA[ + **What it does** + + The tool takes a BioImage.IO model and an image (as TIF or PNG) to be analyzed. The analysis is performed by the model. The model is used to obtain a prediction of the result of the analysis, and the predicted image becomes available as a TIF file in the Galaxy history. + + **Input files** + - BioImage.IO model: Add one of the model from Galaxy file uploader by choosing a "remote" file at "ML Models/bioimaging-models" + - Image to be analyzed: Provide an image as TIF/PNG file + - Provide the necessary input size for the model. This information can be found in the RDF file of each model (RDF file > config > test_information > inputs > size) + + **Output files** + - Predicted image: Predicted image using the BioImage.IO model + - Predicted image matrix: Predicted image matrix in original dimensions + ]]> + </help> + <citations> + <citation type="doi">10.1145/3620665.3640366</citation> + <citation type="doi">10.1101/2022.06.07.495102</citation> + </citations> +</tool>
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/main.py Fri Aug 02 15:40:35 2024 +0000 @@ -0,0 +1,86 @@ +""" +Predict images using AI models from BioImage.IO +""" + +import argparse + +import imageio +import numpy as np +import torch + + +def find_dim_order(user_in_shape, input_image): + """ + Find the correct order of input image's + shape. For a few models, the order of input size + mentioned in the RDF.yaml file is reversed compared + to the input image's original size. If it is reversed, + transpose the image to find correct order of image's + dimensions. + """ + image_shape = list(input_image.shape) + # reverse the input shape provided from RDF.yaml file + correct_order = user_in_shape.split(",")[::-1] + # remove 1s from the original dimensions + correct_order = [int(i) for i in correct_order if i != "1"] + if (correct_order[0] == image_shape[-1]) and (correct_order != image_shape): + input_image = torch.tensor(input_image.transpose()) + return input_image, correct_order + + +if __name__ == "__main__": + arg_parser = argparse.ArgumentParser() + arg_parser.add_argument("-im", "--imaging_model", required=True, help="Input BioImage model") + arg_parser.add_argument("-ii", "--image_file", required=True, help="Input image file") + arg_parser.add_argument("-is", "--image_size", required=True, help="Input image file's size") + + # get argument values + args = vars(arg_parser.parse_args()) + model_path = args["imaging_model"] + input_image_path = args["image_file"] + + # load all embedded images in TIF file + test_data = imageio.v3.imread(input_image_path, index="...") + test_data = np.squeeze(test_data) + test_data = test_data.astype(np.float32) + + # assess the correct dimensions of TIF input image + input_image_shape = args["image_size"] + im_test_data, shape_vals = find_dim_order(input_image_shape, test_data) + + # load model + model = torch.load(model_path) + model.eval() + + # find the number of dimensions required by the model + target_dimension = 0 + for param in model.named_parameters(): + target_dimension = len(param[1].shape) + break + current_dimension = len(list(im_test_data.shape)) + + # update the dimensions of input image if the required image by + # the model is smaller + slices = tuple(slice(0, s_val) for s_val in shape_vals) + + # apply the slices to the reshaped_input + im_test_data = im_test_data[slices] + exp_test_data = torch.tensor(im_test_data) + + # expand input image's dimensions + for i in range(target_dimension - current_dimension): + exp_test_data = torch.unsqueeze(exp_test_data, i) + + # make prediction + pred_data = model(exp_test_data) + pred_data_output = pred_data.detach().numpy() + + # save original image matrix + np.save("output_predicted_image_matrix.npy", pred_data_output) + + # post process predicted file to correctly save as TIF file + pred_data = torch.squeeze(pred_data) + pred_numpy = pred_data.detach().numpy() + + # write predicted TIF image to file + imageio.v3.imwrite("output_predicted_image.tif", pred_numpy, extension=".tif")