initial commit

.gitattributes

@@ -2,7 +2,6 @@
 *.arrow filter=lfs diff=lfs merge=lfs -text
 *.bin filter=lfs diff=lfs merge=lfs -text
 *.bz2 filter=lfs diff=lfs merge=lfs -text
-*.ckpt filter=lfs diff=lfs merge=lfs -text
 *.ftz filter=lfs diff=lfs merge=lfs -text
 *.gz filter=lfs diff=lfs merge=lfs -text
 *.h5 filter=lfs diff=lfs merge=lfs -text
@@ -32,3 +31,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+*.fastai filter=lfs diff=lfs merge=lfs -text

.gitignore

@@ -0,0 +1,147 @@
+# Byte-compiled / optimized / DLL files
+__pycache__/
+*.py[cod]
+*$py.class
+
+# C extensions
+*.so
+
+# Distribution / packaging
+.Python
+build/
+develop-eggs/
+dist/
+downloads/
+eggs/
+.eggs/
+lib/
+lib64/
+parts/
+sdist/
+var/
+wheels/
+share/python-wheels/
+*.egg-info/
+.installed.cfg
+*.egg
+MANIFEST
+
+# PyInstaller
+#  Usually these files are written by a python script from a template
+#  before PyInstaller builds the exe, so as to inject date/other infos into it.
+*.manifest
+*.spec
+
+# Installer logs
+pip-log.txt
+pip-delete-this-directory.txt
+
+# Unit test / coverage reports
+htmlcov/
+.tox/
+.nox/
+.coverage
+.coverage.*
+.cache
+nosetests.xml
+coverage.xml
+*.cover
+*.py,cover
+.hypothesis/
+.pytest_cache/
+cover/
+
+# Translations
+*.mo
+*.pot
+
+# Django stuff:
+*.log
+local_settings.py
+db.sqlite3
+db.sqlite3-journal
+
+# Flask stuff:
+instance/
+.webassets-cache
+
+# Scrapy stuff:
+.scrapy
+
+# Sphinx documentation
+docs/_build/
+
+# PyBuilder
+.pybuilder/
+target/
+
+# Jupyter Notebook
+.ipynb_checkpoints
+
+# IPython
+profile_default/
+ipython_config.py
+
+# pyenv
+#   For a library or package, you might want to ignore these files since the code is
+#   intended to run in multiple environments; otherwise, check them in:
+# .python-version
+
+# pipenv
+#   According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
+#   However, in case of collaboration, if having platform-specific dependencies or dependencies
+#   having no cross-platform support, pipenv may install dependencies that don't work, or not
+#   install all needed dependencies.
+#Pipfile.lock
+
+# PEP 582; used by e.g. github.com/David-OConnor/pyflow
+__pypackages__/
+
+# Celery stuff
+celerybeat-schedule
+celerybeat.pid
+
+# SageMath parsed files
+*.sage.py
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Pycharm
+.idea/
+
+# Spyder project settings
+.spyderproject
+.spyproject
+
+# Rope project settings
+.ropeproject
+
+# mkdocs documentation
+/site
+
+# mypy
+.mypy_cache/
+.dmypy.json
+dmypy.json
+
+# Pyre type checker
+.pyre/
+
+# pytype static type analyzer
+.pytype/
+
+# Cython debug symbols
+cython_debug/
+
+# additionnals stuff
+logs/
+notebooks/
+mlruns/
+flagged/

README.md

@@ -1,10 +1,10 @@
 ---
 title: Mole Analyzer
-emoji: 👀
-colorFrom: gray
-colorTo: purple
+emoji: 🚀
+colorFrom: blue
+colorTo: blue
 sdk: gradio
-sdk_version: 3.16.1
+sdk_version: 3.8.2
 app_file: app.py
 pinned: false
 license: mit

app.py

@@ -0,0 +1,91 @@
+import gradio as gr
+import os
+
+from useful_functions import *
+import useful_functions
+
+f_load_cancer_classifier()
+f_load_cnn_model()
+
+HF_TOKEN = os.getenv('HF_TOKEN')
+hf_writer = gr.HuggingFaceDatasetSaver(HF_TOKEN, "mole-dataset", private=True)
+
+
+def image_classifier(file_path, age, sex, localization):
+    if age == 0:
+        age = 40
+    if sex == "":
+        sex = "unknown"
+    if localization == "":
+        localization = "unknown"
+
+    # file_path = file_path if file_path is not None else file_path_webcam
+
+    preds = f_predict_cnn_with_tta(file_path)
+    label = f_predict_cancer(preds, age, sex, localization)
+    return (dict(zip(useful_functions.lesion_model.dls.vocab, preds)),
+            label)
+
+input_img = gr.Image(tool="editor", type="filepath", source="upload")
+# input_webcam = gr.Image(tool="editor", type="filepath", source="webcam")
+
+input_age = gr.Number(label="age (optionnel)")
+input_sex = gr.Dropdown(label="sex (optionnel)", choices=["male", "female"])
+input_localization = gr.Dropdown(label="localization (optionnel)", choices=["abdomen", "back", "chest", "ear",
+                                              "face", "foot", "genital", "hand",
+                                              "lower extremity", "neck", "scalp", "trunk", "upper extremity"])
+
+output_lesion = gr.Label(label="Lesion detected")
+output_malign = gr.Label(label="Classification")
+
+list_files_examples = os.listdir("examples")
+# examples = [[os.path.join("examples", file), 0, "", ""] for file in list_files_examples if file.endswith("jpg")]
+examples = []
+examples.append([os.path.join("examples", "PXL_20221103_153018529.jpg"), 40, "female", "back"])
+examples.append([os.path.join("examples", "PXL_20221103_153129579.jpg"), 40, "male", "neck"])
+examples.append([os.path.join("examples", "PXL_20221103_153137616.jpg"), 40, "male", "neck"])
+examples.append([os.path.join("examples", "PXL_20221103_153217034.jpg"), 40, "male", "back"])
+examples.append([os.path.join("examples", "PXL_20221103_153256612.jpg"), 40, "male", "upper extremity"])
+examples.append([os.path.join("examples", "ISIC_0025402.jpg"), 70, "male", "lower extremity"])
+
+
+demo = gr.Interface(title="Skin mole analyzer",
+                    description=r"""This is a side project I have been working on to practice working with images.
+                    The purpose is to classify skin lesions (Based on kaggle dataset Skin Cancer MNIST: HAM10000).
+                    The framework used is FastAI/pytorch and the model used is a pre-trained cnn (resnet152). 
+                    I added an extra layer to use age, sex, localization and output of resnet152 to classify the lesion
+                    as suspicious or not (randomForest model).
+                    The lesions detected are the following:
+                    <ul>
+                    <li>Actinic keratoses and intraepithelial carcinoma / Bowen's disease (akiec)</li>
+                    <li>basal cell carcinoma (bcc)</li>
+                    <li>benign keratosis-like lesions (bkl)</li>
+                    <li>dermatofibroma (df)</li>
+                    <li>melanoma (mel)</li>
+                    <li>melanocytic nevi (nv)</li>
+                    <li>vascular lesions (vasc)</li>
+                    </ul>
+                    <b> This is in no case intended as a medical advice, just a pedagogical  exercise.</b> <br />
+                    <i>*Pictures should be relatively well centered on the mole to obtain the best results (cf examples). 
+                    You can use the tools available in the right corner to crop optimally.</i>
+                    """,
+                    article="""[1] Noel Codella, Veronica Rotemberg, Philipp Tschandl, M. Emre Celebi, Stephen Dusza, David Gutman, 
+                            Brian Helba, Aadi Kalloo, Konstantinos Liopyris, Michael Marchetti, Harald Kittler, Allan Halpern: 
+                            "Skin Lesion Analysis Toward Melanoma Detection 2018: A Challenge Hosted by the International Skin Imaging Collaboration (ISIC)",
+                             2018;"<a href="https://arxiv.org/abs/1902.03368">"https://arxiv.org/abs/1902.03368"</a><br />
+                             [2] Tschandl, P., Rosendahl, C. & Kittler, H. The HAM10000 dataset, a large collection of multi-source dermatoscopic 
+                             images of common pigmented skin lesions. Sci. Data 5, 180161 doi:10.1038/sdata.2018.161 (2018).""",
+                    fn=image_classifier,
+                    inputs= [input_img,
+                             input_age,
+                             input_sex,
+                             input_localization],
+                    outputs=[output_lesion,
+                             output_malign],
+                    examples=examples,
+                    allow_flagging="auto",
+                    flagging_options=list(useful_functions.lesion_model.dls.vocab) + ["other"],
+                    flagging_callback=hf_writer
+                    )
+
+demo.launch()

mole_39_dev.yml

@@ -0,0 +1,218 @@
+name: mole_39_dev
+channels:
+  - pytorch
+  - conda-forge
+  - defaults
+dependencies:
+  - anyio=3.5.0=py39haa95532_0
+  - argon2-cffi=21.3.0=pyhd3eb1b0_0
+  - argon2-cffi-bindings=21.2.0=py39h2bbff1b_0
+  - asttokens=2.0.5=pyhd3eb1b0_0
+  - attrs=21.4.0=pyhd3eb1b0_0
+  - babel=2.9.1=pyhd3eb1b0_0
+  - backcall=0.2.0=pyhd3eb1b0_0
+  - beautifulsoup4=4.11.1=py39haa95532_0
+  - blas=2.116=mkl
+  - blas-devel=3.9.0=16_win64_mkl
+  - bleach=4.1.0=pyhd3eb1b0_0
+  - brotlipy=0.7.0=py39h2bbff1b_1003
+  - ca-certificates=2022.6.15.1=h5b45459_0
+  - certifi=2022.6.15.1=pyhd8ed1ab_0
+  - cffi=1.15.1=py39h2bbff1b_0
+  - charset-normalizer=2.0.4=pyhd3eb1b0_0
+  - colorama=0.4.5=py39haa95532_0
+  - cryptography=2.9.2=py39hcd4344a_0
+  - cudatoolkit=11.6.0=hc0ea762_10
+  - debugpy=1.5.1=py39hd77b12b_0
+  - decorator=5.1.1=pyhd3eb1b0_0
+  - defusedxml=0.7.1=pyhd3eb1b0_0
+  - entrypoints=0.4=py39haa95532_0
+  - executing=0.8.3=pyhd3eb1b0_0
+  - freetype=2.10.4=h546665d_1
+  - glib=2.69.1=h5dc1a3c_1
+  - gst-plugins-base=1.18.5=h9e645db_0
+  - gstreamer=1.18.5=hd78058f_0
+  - icu=58.2=ha925a31_3
+  - idna=3.3=pyhd3eb1b0_0
+  - intel-openmp=2022.1.0=h57928b3_3787
+  - ipykernel=6.15.2=py39haa95532_0
+  - ipython=8.4.0=py39haa95532_0
+  - ipython_genutils=0.2.0=pyhd3eb1b0_1
+  - ipywidgets=7.6.5=pyhd3eb1b0_1
+  - jedi=0.18.1=py39haa95532_1
+  - jinja2=3.0.3=pyhd3eb1b0_0
+  - jpeg=9e=h2bbff1b_0
+  - json5=0.9.6=pyhd3eb1b0_0
+  - jsonschema=4.4.0=py39haa95532_0
+  - jupyter=1.0.0=py39haa95532_8
+  - jupyter_client=7.3.5=py39haa95532_0
+  - jupyter_console=6.4.3=pyhd3eb1b0_0
+  - jupyter_core=4.10.0=py39haa95532_0
+  - jupyter_server=1.18.1=py39haa95532_0
+  - jupyterlab=3.4.4=py39haa95532_0
+  - jupyterlab_pygments=0.1.2=py_0
+  - jupyterlab_server=2.12.0=py39haa95532_0
+  - jupyterlab_widgets=1.0.0=pyhd3eb1b0_1
+  - lcms2=2.12=h2a16943_0
+  - libblas=3.9.0=16_win64_mkl
+  - libcblas=3.9.0=16_win64_mkl
+  - libclang=12.0.0=default_h627e005_2
+  - libffi=3.4.2=hd77b12b_4
+  - libiconv=1.16=h2bbff1b_2
+  - liblapack=3.9.0=16_win64_mkl
+  - liblapacke=3.9.0=16_win64_mkl
+  - libogg=1.3.5=h2bbff1b_1
+  - libpng=1.6.37=h2a8f88b_0
+  - libsodium=1.0.18=h62dcd97_0
+  - libtiff=4.2.0=h0c97f57_3
+  - libuv=1.44.2=h8ffe710_0
+  - libvorbis=1.3.7=he774522_0
+  - libwebp=1.2.2=h2bbff1b_0
+  - libxml2=2.9.14=h0ad7f3c_0
+  - libxslt=1.1.35=h2bbff1b_0
+  - lz4-c=1.9.3=h2bbff1b_1
+  - m2w64-gcc-libgfortran=5.3.0=6
+  - m2w64-gcc-libs=5.3.0=7
+  - m2w64-gcc-libs-core=5.3.0=7
+  - m2w64-gmp=6.1.0=2
+  - m2w64-libwinpthread-git=5.0.0.4634.697f757=2
+  - markupsafe=2.1.1=py39h2bbff1b_0
+  - matplotlib-inline=0.1.6=py39haa95532_0
+  - mistune=0.8.4=py39h2bbff1b_1000
+  - mkl=2022.1.0=h6a75c08_874
+  - mkl-devel=2022.1.0=h57928b3_875
+  - mkl-include=2022.1.0=h6a75c08_874
+  - msys2-conda-epoch=20160418=1
+  - nbclassic=0.3.5=pyhd3eb1b0_0
+  - nbclient=0.5.13=py39haa95532_0
+  - nbconvert=6.4.4=py39haa95532_0
+  - nbformat=5.3.0=py39haa95532_0
+  - nest-asyncio=1.5.5=py39haa95532_0
+  - notebook=6.4.12=py39haa95532_0
+  - numpy=1.23.2=py39h1a62c8c_0
+  - olefile=0.46=pyh9f0ad1d_1
+  - openjpeg=2.4.0=hb211442_1
+  - openssl=1.1.1q=h8ffe710_0
+  - packaging=21.3=pyhd3eb1b0_0
+  - pandocfilters=1.5.0=pyhd3eb1b0_0
+  - parso=0.8.3=pyhd3eb1b0_0
+  - pcre=8.45=hd77b12b_0
+  - pickleshare=0.7.5=pyhd3eb1b0_1003
+  - pillow=8.2.0=py39h1a9d4f7_1
+  - pip=22.1.2=py39haa95532_0
+  - ply=3.11=py39haa95532_0
+  - prometheus_client=0.14.1=py39haa95532_0
+  - prompt-toolkit=3.0.20=pyhd3eb1b0_0
+  - prompt_toolkit=3.0.20=hd3eb1b0_0
+  - psutil=5.9.0=py39h2bbff1b_0
+  - pure_eval=0.2.2=pyhd3eb1b0_0
+  - pycparser=2.21=pyhd3eb1b0_0
+  - pygments=2.11.2=pyhd3eb1b0_0
+  - pyopenssl=20.0.1=pyhd3eb1b0_1
+  - pyparsing=3.0.9=py39haa95532_0
+  - pyqt=5.15.7=py39hd77b12b_0
+  - pyqt5-sip=12.11.0=py39hd77b12b_0
+  - pyrsistent=0.18.0=py39h196d8e1_0
+  - pysocks=1.7.1=py39haa95532_0
+  - python=3.9.13=h6244533_1
+  - python-dateutil=2.8.2=pyhd3eb1b0_0
+  - python-fastjsonschema=2.16.2=py39haa95532_0
+  - python_abi=3.9=2_cp39
+  - pytz=2022.1=py39haa95532_0
+  - pywinpty=2.0.2=py39h5da7b33_0
+  - pyzmq=23.2.0=py39hd77b12b_0
+  - qt-main=5.15.2=he8e5bd7_7
+  - qt-webengine=5.15.9=hb9a9bb5_4
+  - qtconsole=5.3.2=py39haa95532_0
+  - qtpy=2.2.0=py39haa95532_0
+  - qtwebkit=5.212=h3ad3cdb_4
+  - requests=2.28.1=py39haa95532_0
+  - send2trash=1.8.0=pyhd3eb1b0_1
+  - setuptools=63.4.1=py39haa95532_0
+  - sip=6.6.2=py39hd77b12b_0
+  - six=1.16.0=pyhd3eb1b0_1
+  - sniffio=1.2.0=py39haa95532_1
+  - soupsieve=2.3.1=pyhd3eb1b0_0
+  - sqlite=3.39.2=h2bbff1b_0
+  - stack_data=0.2.0=pyhd3eb1b0_0
+  - tbb=2021.5.0=h2d74725_1
+  - terminado=0.13.1=py39haa95532_0
+  - testpath=0.6.0=py39haa95532_0
+  - tk=8.6.12=h8ffe710_0
+  - toml=0.10.2=pyhd3eb1b0_0
+  - tornado=6.2=py39h2bbff1b_0
+  - traitlets=5.1.1=pyhd3eb1b0_0
+  - typing-extensions=4.3.0=py39haa95532_0
+  - typing_extensions=4.3.0=py39haa95532_0
+  - tzdata=2022a=hda174b7_0
+  - urllib3=1.26.11=py39haa95532_0
+  - vc=14.2=h21ff451_1
+  - vs2015_runtime=14.27.29016=h5e58377_2
+  - wcwidth=0.2.5=pyhd3eb1b0_0
+  - webencodings=0.5.1=py39haa95532_1
+  - websocket-client=0.58.0=py39haa95532_4
+  - wheel=0.37.1=pyhd3eb1b0_0
+  - widgetsnbextension=3.5.2=py39haa95532_0
+  - win_inet_pton=1.1.0=py39haa95532_0
+  - wincertstore=0.2=py39haa95532_2
+  - winpty=0.4.3=4
+  - xz=5.2.6=h8d14728_0
+  - zeromq=4.3.4=hd77b12b_0
+  - zlib=1.2.12=h8cc25b3_3
+  - zstd=1.5.2=h19a0ad4_0
+  - pip:
+    - alembic==1.8.1
+    - blis==0.9.1
+    - catalogue==2.0.8
+    - click==8.1.3
+    - cloudpickle==2.2.0
+    - confection==0.0.1
+    - cycler==0.11.0
+    - cymem==2.0.6
+    - databricks-cli==0.17.3
+    - docker==5.0.3
+    - fastai==2.7.9
+    - fastcore==1.5.26
+    - fastdownload==0.0.7
+    - fastprogress==1.0.3
+    - flask==2.2.2
+    - fonttools==4.37.1
+    - gitdb==4.0.9
+    - gitpython==3.1.27
+    - greenlet==1.1.3
+    - importlib-metadata==4.12.0
+    - itsdangerous==2.1.2
+    - kiwisolver==1.4.4
+    - langcodes==3.3.0
+    - mako==1.2.2
+    - matplotlib==3.5.3
+    - mlflow==1.28.0
+    - murmurhash==1.0.8
+    - oauthlib==3.2.1
+    - pathy==0.6.2
+    - preshed==3.0.7
+    - prometheus-flask-exporter==0.20.3
+    - protobuf==4.21.5
+    - pydantic==1.9.2
+    - pyjwt==2.4.0
+    - pywin32==227
+    - pyyaml==6.0
+    - querystring-parser==1.2.4
+    - smart-open==5.2.1
+    - smmap==5.0.0
+    - spacy==3.4.1
+    - spacy-legacy==3.0.10
+    - spacy-loggers==1.0.3
+    - sqlalchemy==1.4.41
+    - sqlparse==0.4.2
+    - srsly==2.4.4
+    - tabulate==0.8.10
+    - thinc==8.1.1
+    - torchviz==0.0.2
+    - tqdm==4.64.1
+    - typer==0.4.2
+    - waitress==2.1.2
+    - wasabi==0.10.1
+    - werkzeug==2.2.2
+    - zipp==3.8.1
+prefix: C:\Users\jeanpoll\.conda\envs\mole_39_dev

mole_models/ed52a28a7b504ff7ba851c850221d1dd/MLmodel

@@ -0,0 +1,13 @@
+flavors:
+  fastai:
+    code: null
+    data: model.fastai
+    fastai_version: 2.7.9
+  python_function:
+    data: model.fastai
+    env: conda.yaml
+    loader_module: mlflow.fastai
+    python_version: 3.9.13
+mlflow_version: 1.28.0
+model_uuid: 7811b4341435426287805c26e84e31d6
+utc_time_created: '2022-10-22 17:09:31.990516'

mole_models/ed52a28a7b504ff7ba851c850221d1dd/conda.yaml

@@ -0,0 +1,9 @@
+channels:
+- conda-forge
+dependencies:
+- python=3.9.13
+- pip<=22.1.2
+- pip:
+  - mlflow
+  - fastai==2.7.9
+name: mlflow-env

mole_models/ed52a28a7b504ff7ba851c850221d1dd/model.fastai

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96db40fa4c0819530b82edee1cc58725057fa05b1b9679d5b9fc0b6d5da852f1
+size 244863906

mole_models/ed52a28a7b504ff7ba851c850221d1dd/python_env.yaml

@@ -0,0 +1,7 @@
+python: 3.9.13
+build_dependencies:
+- pip==22.1.2
+- setuptools==63.4.1
+- wheel==0.37.1
+dependencies:
+- -r requirements.txt

mole_models/ed52a28a7b504ff7ba851c850221d1dd/requirements.txt

@@ -0,0 +1,2 @@
+mlflow
+fastai==2.7.9

requirements.txt

@@ -0,0 +1,9 @@
+numpy
+pandas==1.5.1
+torch==1.12.1
+scikit-learn==1.1.3
+gradio==3.15
+fastai==2.7.9
+mlflow==1.28.0
+huggingface_hub==0.10.1
+

useful_functions.py

@@ -0,0 +1,71 @@
+import mlflow
+import pandas as pd
+import joblib
+import os
+
+from fastai.data.block import MultiCategoryBlock, RandomSplitter, DataBlock, CategoryBlock
+from fastai.vision.data import *
+from fastai.vision.learner import *
+from fastai.vision.all import *
+import pathlib
+plt = platform.system()
+if plt == 'Linux': pathlib.WindowsPath = pathlib.PosixPath
+
+
+lesion_model = None
+pipe_cancer = None
+
+
+def f_load_cnn_model():
+    global lesion_model
+    logged_model = os.path.join("mole_models", "ed52a28a7b504ff7ba851c850221d1dd")
+    lesion_model = mlflow.fastai.load_model(logged_model)
+    lesion_model.cbs.remove(lesion_model.cbs[6])
+    lesion_model.cbs.remove(lesion_model.cbs[4])
+    lesion_model.cbs.remove(lesion_model.cbs[3])
+
+
+def get_image_files(df):
+    # df = df.assign(path=lambda x: working_dataset_path + x["path"])
+    return df
+
+def get_x(df):
+    return df["path"]
+
+def get_y(df):
+    return df["dx"]
+
+
+def f_create_df_with_files_input(file_path):
+    list_files = [file_path]
+    list_labels = [""]
+    df = pd.DataFrame(data={"path": list_files, "dx": list_labels})
+    return df
+
+
+def f_predict_cnn_with_tta(file_path):
+    df = f_create_df_with_files_input(file_path)
+    dl = lesion_model.dls.test_dl(df)
+    dl.after_item = Pipeline([ToTensor, Resize(700, method=ResizeMethod.Crop), RandomResizedCrop(350)])
+    pred, _targ = lesion_model.tta(dl=dl, n=4, use_max=False)
+    return pred.tolist()[0]
+
+
+def f_predict_cnn_simple(file_path):
+    preds = lesion_model.predict(file_path)
+    return preds[2].tolist()
+
+
+def f_load_cancer_classifier():
+    global pipe_cancer
+    pipe_cancer = joblib.load(os.path.join("mole_models", "pipeline.rf_classifier_mole"))
+
+def f_predict_cancer(preds, age, sex, localization):
+    df = pd.DataFrame(data=[preds], columns=lesion_model.dls.vocab)
+    df["age"] = age
+    df["sex"] = sex
+    df["localization"] = localization
+    df["label"] = ""
+    label = pipe_cancer.predict(df)[0]
+    label = "Possibly suspicious" if label=="cancer" else "benign"
+    return label