Skip to content

Commit

Permalink
Adding new/updated DL4MicEverywhere_embedseg-2d-zerocostdl4mic_1.15.0
Browse files Browse the repository at this point in the history
IvanHCenalmor committed Oct 15, 2024
1 parent 6fc1d65 commit e0656f1
Showing 4 changed files with 206 additions and 0 deletions.
Binary file modified album_catalog_index.db
Binary file not shown.
Original file line number Diff line number Diff line change
@@ -0,0 +1,8 @@
# Changelog
All notable changes to this project will be documented in this file.

The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).

## [1.15.0] - 2024-10-15
../CHANGELOG.md
164 changes: 164 additions & 0 deletions solutions/DL4MicEverywhere/embedseg-2d-zerocostdl4mic/solution.py
Original file line number Diff line number Diff line change
@@ -0,0 +1,164 @@
###album catalog: cellcanvas

# Based on https://github.com/HenriquesLab/DL4MicEverywhere/blob/main/notebooks/ZeroCostDL4Mic_notebooks/Embedseg_DL4Mic/configuration.yaml
# and https://github.com/betaseg/solutions/blob/main/solutions/io.github.betaseg/cellsketch-plot/solution.py

from album.runner.api import setup
import subprocess

try:
subprocess.check_output('nvidia-smi')
gpu_access = True
except Exception:
gpu_access = False

def install():
from album.runner.api import get_app_path
from git import Repo
import subprocess
import requests
import shutil
import os

# Clone the DL4MicEverywhere repository
clone_url = "https://github.com/HenriquesLab/DL4MicEverywhere"
repo_path = get_app_path().joinpath("DL4MicEverywhere")
Repo.clone_from(clone_url, repo_path)
assert (repo_path.exists())

# URL of the notebook you want to download
notebook_url = "https://raw.githubusercontent.com/HenriquesLab/ZeroCostDL4Mic/master/Colab_notebooks/EmbedSeg_2D_ZeroCostDL4Mic.ipynb"

notebook_path = get_app_path().joinpath("EmbedSeg_2D_ZeroCostDL4Mic.ipynb")
notebook_path.parent.mkdir(parents=True, exist_ok=True)

response = requests.get(notebook_url)
response.raise_for_status()

with open(notebook_path, 'wb') as notebook_file:
notebook_file.write(response.content)

assert notebook_path.exists(), "Notebook download failed"

# Convert the notebook to its colabless form
section_to_remove = "1.0. 1.1. 2."
section_to_remove = section_to_remove.split(' ')

python_command = ["python", ".tools/notebook_autoconversion/transform.py", "-p", f"{get_app_path()}", "-n", "EmbedSeg_2D_ZeroCostDL4Mic.ipynb", "-s"]
python_command += section_to_remove

subprocess.run(python_command, cwd=to)
subprocess.run(["mv", get_app_path().joinpath("colabless_EmbedSeg_2D_ZeroCostDL4Mic.ipynb"), get_app_path().joinpath("EmbedSeg_2D_ZeroCostDL4Mic.ipynb")])

# Remove the cloned DL4MicEverywhere repository
if os.name == 'nt':
os.system(f'rmdir /s /q "{to}"')
else:
# rmtree has no permission to do this on Windows
shutil.rmtree(to)

def run():
from album.runner.api import get_args, get_app_path
import subprocess
import os

# Fetch arguments and paths
args = get_args()
app_path = get_app_path()

# Path to the downloaded notebook
notebook_path = app_path.joinpath("EmbedSeg_2D_ZeroCostDL4Mic.ipynb")

# Ensure the notebook exists
assert notebook_path.exists(), "Notebook does not exist"

# Output path for running the notebook
output_path = args.path
os.makedirs(output_path, exist_ok=True)
print(f"Saving output to {output_path}")

# Set the LD_LIBRARY_PATH to allow TensorFlow to find the CUDA libraries
global gpu_access
if gpu_access:
os.environ["LD_LIBRARY_PATH"] = f"{os.environ['LD_LIBRARY_PATH']}:{os.environ['CONDA_PREFIX']}/lib"

# Optionally, launch the Jupyter notebook to show the results
subprocess.run(["jupyter", "lab", str(notebook_path)], cwd=str(output_path))

if gpu_access:
channels = """
- conda-forge
- nvidia
- anaconda
- defaults
"""
dependencies = """
- python=3.9
- cudatoolkit=11.7.1
- cudnn=8.5.0
- pip
- pkg-config
"""
else:
channels = """
- conda-forge
- defaults
"""
dependencies = f"""
- python=3.9
- pip
- pkg-config
"""

env_file = f"""
channels:
{channels}
dependencies:
{dependencies}
- pip:
- GitPython==3.1.43
- astropy==6.0.0
- torch<=1.9.0
- torchvision<=0.10.0
- scikit-learn==1.3.2
- embedseg==0.2.3
- wget
- fpdf
- h5py
- tabulate
- matplotlib==3.5.0
- scipy
- tifffile
- numba==0.56
- tqdm
- jupyter
- pandas==1.3.0
- seaborn
- scikit-image
- colorspacious
- pytest
- pycocotools
"""

setup(
group="DL4MicEverywhere",
name="embedseg-2d-zerocostdl4mic",
version="1.15.0",
solution_creators=["DL4Mic team", "album team"],
title="embedseg-2d-zerocostdl4mic implementation.",
description="Instance segmentation of 2D images. EmbedSeg 2D is a deep-learning method that can be used to segment object from bioimages and was first published by Lalit et al. in 2021, on arXiv.",
documentation="https://raw.githubusercontent.com/HenriquesLab/ZeroCostDL4Mic/master/BioimageModelZoo/README.md",
tags=['colab', 'notebook', 'EmbedSeg', 'Segmentation', 'ZeroCostDL4Mic', 'dl4miceverywhere'],
args=[{
"name": "path",
"type": "string",
"default": ".",
"description": "What is your working path?"
}],
cite=[{'doi': 'https://doi.org/10.1038/s41467-021-22518-0', 'text': 'von Chamier, L., Laine, R.F., Jukkala, J. et al. Democratising deep learning for microscopy with ZeroCostDL4Mic. Nat Commun 12, 2276 (2021). https://doi.org/10.1038/s41467-021-22518-0'}, {'text': 'Manan Lalit, Pavel Tomancak, Florian Jug. Embedding-based Instance Segmentation in Microscopy. arXiv:2101.10033.', 'url': 'https://arxiv.org/abs/2101.10033'}],
album_api_version="0.5.1",
covers=[],
run=run,
install=install,
dependencies={"environment_file": env_file},
)
34 changes: 34 additions & 0 deletions solutions/DL4MicEverywhere/embedseg-2d-zerocostdl4mic/solution.yml
Original file line number Diff line number Diff line change
@@ -0,0 +1,34 @@
album_api_version: 0.5.1
args:
- default: .
description: What is your working path?
name: path
type: string
changelog: ../CHANGELOG.md
cite:
- doi: https://doi.org/10.1038/s41467-021-22518-0
text: von Chamier, L., Laine, R.F., Jukkala, J. et al. Democratising deep learning
for microscopy with ZeroCostDL4Mic. Nat Commun 12, 2276 (2021). https://doi.org/10.1038/s41467-021-22518-0
- text: Manan Lalit, Pavel Tomancak, Florian Jug. Embedding-based Instance Segmentation
in Microscopy. arXiv:2101.10033.
url: https://arxiv.org/abs/2101.10033
covers: []
description: Instance segmentation of 2D images. EmbedSeg 2D is a deep-learning method
that can be used to segment object from bioimages and was first published by Lalit
et al. in 2021, on arXiv.
documentation: https://raw.githubusercontent.com/HenriquesLab/ZeroCostDL4Mic/master/BioimageModelZoo/README.md
group: DL4MicEverywhere
name: embedseg-2d-zerocostdl4mic
solution_creators:
- DL4Mic team
- album team
tags:
- colab
- notebook
- EmbedSeg
- Segmentation
- ZeroCostDL4Mic
- dl4miceverywhere
timestamp: '2024-10-15T17:48:35.346227'
title: embedseg-2d-zerocostdl4mic implementation.
version: 1.15.0

0 comments on commit e0656f1

Please sign in to comment.