|
| 1 | +import os |
| 2 | +import subprocess |
| 3 | +from shutil import copyfile |
| 4 | + |
| 5 | +import imageio.v3 as imageio |
| 6 | +import napari |
| 7 | +import pandas as pd |
| 8 | +import zarr |
| 9 | +from flamingo_tools.test_data import _sample_registry |
| 10 | + |
| 11 | +data_dict = { |
| 12 | + "SGN": "PV", |
| 13 | + "IHC": "VGlut3", |
| 14 | + "SGN-lowres": "PV-lowres", |
| 15 | + "IHC-lowres": "MYO-lowres", |
| 16 | + "Synapses": "CTBP2", |
| 17 | +} |
| 18 | + |
| 19 | + |
| 20 | +def check_segmentation_model(model_name): |
| 21 | + output_folder = f"result_{model_name}" |
| 22 | + os.makedirs(output_folder, exist_ok=True) |
| 23 | + input_path = os.path.join(output_folder, f"{model_name}.tif") |
| 24 | + if not os.path.exists(input_path): |
| 25 | + data_path = _sample_registry().fetch(data_dict[model_name]) |
| 26 | + copyfile(data_path, input_path) |
| 27 | + |
| 28 | + subprocess.run( |
| 29 | + ["flamingo_tools.run_segmentation", "-i", input_path, "-o", output_folder, "-m", model_name] |
| 30 | + ) |
| 31 | + output_path = os.path.join(output_folder, "segmentation.zarr") |
| 32 | + segmentation = zarr.open(output_path)["segmentation"][:] |
| 33 | + |
| 34 | + image = imageio.imread(input_path) |
| 35 | + v = napari.Viewer() |
| 36 | + v.add_image(image) |
| 37 | + v.add_labels(segmentation, name=f"{model_name}-segmentation") |
| 38 | + napari.run() |
| 39 | + |
| 40 | + |
| 41 | +def check_detection_model(): |
| 42 | + model_name = "Synapses" |
| 43 | + output_folder = f"result_{model_name}" |
| 44 | + os.makedirs(output_folder, exist_ok=True) |
| 45 | + input_path = os.path.join(output_folder, f"{model_name}.tif") |
| 46 | + if not os.path.exists(input_path): |
| 47 | + data_path = _sample_registry().fetch(data_dict[model_name]) |
| 48 | + copyfile(data_path, input_path) |
| 49 | + |
| 50 | + subprocess.run( |
| 51 | + ["flamingo_tools.run_detection", "-i", input_path, "-o", output_folder, "-m", model_name] |
| 52 | + ) |
| 53 | + output_path = os.path.join(output_folder, "synapse_detection.tsv") |
| 54 | + prediction = pd.read_csv(output_path, sep="\t")[["z", "y", "x"]] |
| 55 | + |
| 56 | + image = imageio.imread(input_path) |
| 57 | + v = napari.Viewer() |
| 58 | + v.add_image(image) |
| 59 | + v.add_points(prediction) |
| 60 | + napari.run() |
| 61 | + |
| 62 | + |
| 63 | +def main(): |
| 64 | + # SGN segmentation: |
| 65 | + # - Prediction works well on the CPU. |
| 66 | + # check_segmentation_model("SGN") |
| 67 | + |
| 68 | + # IHC segmentation: |
| 69 | + # - Prediction does not work well on the CPU. |
| 70 | + # check_segmentation_model("IHC") |
| 71 | + |
| 72 | + # SGN segmentation (lowres): |
| 73 | + # - Prediction does not work well on the CPU. |
| 74 | + # check_segmentation_model("SGN-lowres") |
| 75 | + |
| 76 | + # IHC segmentation (lowres): |
| 77 | + # - The prediction seems to work (on the CPU), but a lot of merges. |
| 78 | + # -> Update the segmentation params? |
| 79 | + # check_segmentation_model("IHC-lowres") |
| 80 | + |
| 81 | + # Synapse detection: |
| 82 | + # - Prediction works well on the CPU. |
| 83 | + check_detection_model() |
| 84 | + |
| 85 | + |
| 86 | +if __name__ == "__main__": |
| 87 | + main() |
0 commit comments