Skinet v1.0

LICENSE

@@ -0,0 +1,23 @@
+Skinet (Segmentation of the Kidney through a Neural nETwork) Project
+
+MIT License
+
+Copyright (c) 2020 Skinet Team
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

LICENSE_MATTERPORT

@@ -0,0 +1,23 @@
+Mask R-CNN
+
+The MIT License (MIT)
+
+Copyright (c) 2017 Matterport, Inc.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.

README.md

@@ -0,0 +1,12 @@
+# SKINET (Segmentation of the KIdney through a Neural nETwork) Project
+
+SKINET Project is meant to perform a segmentation of a kidney's biopsy or a nephrectomy and recognize the different histological structures. By doing that, it is possible to analyze kidneys more precisely and get a better understanding of their behaviors.
+
+The project's code is based on [Matterport's Mask R-CNN](https://github.com/matterport/Mask_RCNN) and [Navidyou's repository](https://github.com/navidyou/Mask-RCNN-implementation-for-cell-nucleus-detection-executable-on-google-colab-).
+
+This project is a collaboration between a Nephrology team from [Dijon Burgundy Teaching Hospital](https://www.chu-dijon.fr/), [LEAD Laboratory](http://leadserv.u-bourgogne.fr/en/), and a student from [ESIREM](https://esirem.u-bourgogne.fr/), all located in Dijon, Burgundy, France.
+
+## Inference tool
+Last : [![Open Inference Tool In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/SkinetTeam/Skinet/blob/main/Skinet_Inference_Tool.ipynb)
+
+v1.0 : [![Open Inference Tool In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/SkinetTeam/Skinet/blob/v1.0/Skinet_Inference_Tool.ipynb)

common_utils.py

@@ -0,0 +1,141 @@
+"""
+Skinet (Segmentation of the Kidney through a Neural nETwork) Project
+Common display/math methods
+
+Copyright (c) 2021 Skinet Team
+Licensed under the MIT License (see LICENSE for details)
+Written by Adrien JAUGEY
+"""
+from datetime import datetime
+import numpy as np
+
+
+def progressBar(value, maxValue, prefix="", suffix="", forceNewLine=False, size=20, full='█', cursor='▒',
+                empty='░'):
+    """
+    Prints a progress bar
+    Based on https://stackoverflow.com/questions/6169217/replace-console-output-in-python
+    :param value: the current progress value
+    :param maxValue: the maximum value that could be given
+    :param prefix: text to display before the progress bar
+    :param suffix: text to display after the progress bar
+    :param forceNewLine: False by default, if True a new line will be created even if not at the end
+    :param size: size of the bar itself
+    :param full: the character to use for the completed part of the bar
+    :param cursor: the character to use for the current position
+    :param empty: the character to use for the empty part of the bar
+    :return: None
+    """
+    percent = float(value) / maxValue
+    nbFullChar = int(percent * size)
+    bar = full * nbFullChar + (cursor if percent > 0 and nbFullChar < size else "")
+    emptyBar = empty * (size - len(bar))
+    print(f'\r{prefix} {bar}{emptyBar} {percent: 6.2%} {suffix}',
+          end='\n' if value == maxValue or forceNewLine else "", flush=True)
+
+
+def progressText(value, maxValue, onlyRaw=False, onlyPercent=False):
+    if onlyRaw and onlyPercent:
+        return None
+    elif onlyRaw:
+        return f"({value}/{maxValue})"
+    else:
+        percent = float(value) / maxValue
+        if onlyPercent:
+            return f"({percent:06.2%})"
+        else:
+            return f"({value}/{maxValue} | {percent:06.2%})"
+
+
+def formatTime(seconds: int, minutes: int = 0, hours: int = 0):
+    """
+    Returns a string representing the given time
+    :param seconds: number of seconds
+    :param minutes: if given, number of minutes adding to seconds
+    :param hours:  if given, number of hours adding to seconds
+    :return: formated time as string of "hh:mm:ss" format
+    """
+    seconds += minutes * 60 + hours * 3600
+    seconds = int(seconds)
+    h = seconds // 3600
+    m = (seconds % 3600) // 60
+    s = seconds % 60
+    hText = f"{h:02d}:" if h != 0 else ""
+    mText = f"{m:02d}:" if h + m != 0 else ""
+    sText = f"{s:02d}" if h + m != 0 else f"{s:02d}s"
+    return f"{hText}{mText}{sText}"
+
+
+def formatDate(date: datetime = None, dateOnly=False, timeOnly=False):
+    """
+    Returns date as string
+    :param date: datetime.datetime object to use specific date, current date and time by default
+    :param dateOnly: if True, will only display the date, not the time
+    :param timeOnly: if True, will only display the time, not the date
+    :return: formatted date as string
+    """
+    if date is None:
+        date = datetime.now()
+    else:
+        assert type(date) is datetime, "Provide date as datetime.datetime object"
+    dateFormat = '%Y-%m-%d'
+    timeFormat = '%H-%M-%S'
+    if dateOnly and timeOnly:
+        outputFormat = ""
+    elif dateOnly:
+        outputFormat = dateFormat
+    elif timeOnly:
+        outputFormat = timeFormat
+    else:
+        outputFormat = f"{dateFormat}_{timeFormat}"
+    return date.strftime(outputFormat)
+
+
+def format_number(num, maxLength=None):
+    """
+    Formats a number using a metric prefix such as K (Kilo), M (Mega)... up to Y (Yotta)
+    :param num: the number to format
+    :param maxLength: maximum length of the formatted number, will reduce number of decimals from 2 to 0 depending on
+                      the length but will at least return abs(num / 1000^N) followed by the metric prefix for 1000^N.
+    :return: the formatted number with up to 2 decimal digits
+    """
+    # https://stackoverflow.com/questions/579310/formatting-long-numbers-as-strings-in-python
+    suffixes = ['', 'K', 'M', 'G', 'T', 'P', 'E', 'Z', 'Y']
+    magnitude = 0
+    isInt = np.issubdtype(type(num), np.integer)
+    while abs(num) >= 1000 and magnitude < len(suffixes) - 1:
+        magnitude += 1
+        num /= 1000.0
+
+    if magnitude == 0 and isInt:
+        return str(int(num))
+
+    if maxLength is not None:
+        possibleText = [f"{num:.2f}{suffixes[magnitude]}",
+                        f"{num:.1f}{suffixes[magnitude]}",
+                        f"{num:.0f}{suffixes[magnitude]}"]
+        for text in possibleText:
+            if len(text) <= maxLength:
+                return text
+        return possibleText[-1]
+    return f"{num:.1f}{suffixes[magnitude]}"
+
+
+def combination(setSize, combinationSize):
+    """
+    Computes the number of k-combinations in a set
+    Source : https://python.jpvweb.com/python/mesrecettespython/doku.php?id=combinaisons
+    :param setSize: number of elements in the set
+    :param combinationSize: number of elements in a combination
+    :return: number of k-combinations
+    """
+    if combinationSize > setSize // 2:
+        combinationSize = setSize - combinationSize
+    x = 1
+    y = 1
+    i = setSize - combinationSize + 1
+    while i <= setSize:
+        x = (x * i) // y
+        y += 1
+        i += 1
+    return x

environment.yml

@@ -0,0 +1,25 @@
+name: Skinet
+channels:
+  - defaults
+dependencies:
+  - python=3.7
+  - pip:
+    - tensorflow-gpu==1.15.5
+    - tensorboard==1.15
+    - keras==2.3.1
+    - h5py==2.10.0
+    - jupyter
+    - matplotlib
+    - numpy==1.18.3
+    - opencv-python==4.2.0.34
+    - imgaug==0.4.0
+    - ipython
+    - jsonschema
+    - html5lib
+    - imageio==2.8.0
+    - imagesize==1.2.0
+    - pillow==7.1.1
+    - scikit-image==0.15.0
+    - scipy==1.1.0
+
+

mrcnn/compat.py

@@ -0,0 +1,78 @@
+"""
+Skinet (Segmentation of the Kidney through a Neural nETwork) Project
+
+Copyright (c) 2021 Skinet Team
+Licensed under the MIT License (see LICENSE for details)
+Written by Adrien JAUGEY
+"""
+import tensorflow as tf
+
+version = tf.__version__
+TF_MAJOR, TF_MINOR, TF_PATCH = [int(v) for v in version.split('.')][:3]
+
+
+def get_version():
+    return TF_MAJOR, TF_MINOR, TF_PATCH
+
+
+############################################################
+#  Arguments rewrite
+############################################################
+def crop_and_resize_v1(image, boxes, box_indices, crop_size, method, extrapolation_value, name):
+    return tf.image.crop_and_resize(image, boxes, box_ind=box_indices, crop_size=crop_size, method=method,
+                                    extrapolation_value=extrapolation_value, name=name)
+
+
+def crop_and_resize_v2(image, boxes, box_indices, crop_size, method, extrapolation_value, name):
+    return tf.image.crop_and_resize(image, boxes, box_indices=box_indices, crop_size=crop_size, method=method,
+                                    extrapolation_value=extrapolation_value, name=name)
+
+
+############################################################
+#  Defining methods to use
+############################################################
+if TF_MAJOR == 1 and 3 <= TF_MINOR <= 15:
+    if TF_MINOR < 14:
+        WHERE_FUNC = tf.where
+    else:
+        WHERE_FUNC = tf.compat.v1.where_v2
+
+    if TF_MINOR < 13:
+        CROP_AND_RESIZE_FUNC = crop_and_resize_v1
+        INTERSECTION_FUNC = tf.sets.set_intersection
+    else:  # TF >= 1.13
+        CROP_AND_RESIZE_FUNC = crop_and_resize_v2
+        INTERSECTION_FUNC = tf.sets.intersection
+
+    if TF_MINOR < 12:
+        TO_DENSE_FUNC = tf.sparse_tensor_to_dense
+    else:  # TF >= 1.12
+        TO_DENSE_FUNC = tf.sparse.to_dense
+
+    if TF_MINOR < 10:
+        LOG_FUNC = tf.log
+    else:  # TF >= 1.10
+        LOG_FUNC = tf.math.log
+else:
+    raise NotImplementedError(f"Compatibility with TF {tf.__version__} is not implemented")
+
+
+def crop_and_resize(image, boxes, box_indices=None, crop_size=None, method='bilinear', extrapolation_value=0,
+                    name=None):
+    return CROP_AND_RESIZE_FUNC(image, boxes, box_indices, crop_size, method, extrapolation_value, name)
+
+
+def intersection(a, b, validate_indices=True):
+    return INTERSECTION_FUNC(a=a, b=b, validate_indices=validate_indices)
+
+
+def log(x, name=None):
+    return LOG_FUNC(x, name=name)
+
+
+def to_dense(sp_input, default_value=None, validate_indices=True, name=None):
+    return TO_DENSE_FUNC(sp_input=sp_input, default_value=default_value, validate_indices=validate_indices, name=name)
+
+
+def where(condition, x=None, y=None, name=None):
+    return WHERE_FUNC(condition=condition, x=x, y=y, name=name)