Commit 254de9e7 authored by Nicolas Elie's avatar Nicolas Elie
Browse files

pour model

parent 6fe2b1ab

.ipynb_checkpoints/JSON_Qupath_to_ImageLabel-checkpoint.ipynb

0 → 100755
+135 −0
Changes for .ipynb_checkpoints/JSON_Qupath_to_ImageLabel-checkpoint.ipynb: 135 added lines, 0 removed lines.
Original line number Diff line number Diff line 
%% Cell type:code id: tags:



``` python

from __future__ import print_function

from collections import OrderedDict

from io import BytesIO

import random

import struct

import gzip

import geojson

from shapely.geometry import shape

from shapely.strtree import STRtree

from shapely.geometry import Point

from shapely.geometry import Polygon

from skimage.draw import polygon

import numpy

import multiprocessing as mp

import subprocess, os, skimage,mahotas, cv2

from skimage import io, filters, morphology, util, transform, segmentation, measure,io,feature,exposure

import math

#from skimage.segmentation import clear_border

import pandas as pd

import glob, scipy

import itertools

from skimage.filters.rank import entropy

import javabridge

import bioformats

from xml import etree as et

```



%% Cell type:code id: tags:



``` python

def drawnucleiFinal(SizeImage,allshape):

    """

    SizeImage : size image

    allshape : coordinates of objects

    """

    TotalUnique=numpy.zeros((SizeImage[3],SizeImage[2]),numpy.uint16())



    compteurU=1

    Dictionnaire={}

    for objet in allshape:

        ROI=numpy.zeros((SizeImage[3],SizeImage[2]),numpy.uint8())

        if objet ["geometry"]["type"]=="Polygon":

            # Create a Polygon from the coordinates

            poly = Polygon(objet["geometry"]["coordinates"][0])



            "conversion du polygon en numarray arrondi"

            contour=numpy.uint32(numpy.around(numpy.asarray(poly.exterior.coords)))



            "creation de la matrice de soustraction pour revenir à un depart à 0,0"

            try:

                bc_arr, bc_row_means = numpy.broadcast_arrays(contour, numpy.array([SizeImage[0],SizeImage[1]]))

            except:

                print('erreur')

            polygone=bc_arr-bc_row_means

            "dessin du polygon sur l'image pour faire un masque"

            c,r=numpy.split(polygone,2,axis=1)



            rr, cc = polygon(r.reshape(r.shape[0]), c.reshape(c.shape[0]))

            tab=numpy.where(rr>=SizeImage[3])

            rr[tab]=SizeImage[3]-1



            tab=numpy.where(cc>=SizeImage[2])

            cc[tab]=SizeImage[2]-1



            try:

                ROI[rr,cc]=255

                Tab=numpy.where(ROI==255)

                if objet["properties"]["classification"]["name"]=="sombre":

                    Dictionnaire[compteurU]=1

                    TotalUnique[Tab]=compteurU

                    compteurU+=1



                if objet["properties"]["classification"]["name"]=="clair":

                    Dictionnaire[compteurU]=2

                    TotalUnique[Tab]=compteurU

                    compteurU+=1

            except :

                print("erreur 2")

                pass



    return TotalUnique,Dictionnaire

```



%% Cell type:code id: tags:



``` python

"dossier contenant les contours dessinées sur qupath et exportés en json"

Liste=sorted(glob.glob("Examples/qupathProject/ExportJSON/*.json"))



"dossier des images d'origines"

dossierimage="Examples/"



"parcours des fichiers de contours"

for json_detections in Liste:

        namefile=json_detections[:-5]

        #namefiletif=json_detections[:-5].split("/")[-1]+".tif"

        namefiletif=json_detections[:-19].split("/")[-1]

        with open(json_detections) as g:

                alldetections = geojson.load(g)



        imagecouleur=io.imread(dossierimage+namefiletif)

        "Contour design, labeling and dictionary creation for classification"

        TotalUnique,my_dict=drawnucleiFinal([0,0,1544,1038],alldetections)



        "lecture de l'image couleur d'origine"

        imagecouleur=io.imread(dossierimage+namefiletif)

        nombre=0

        "decoupage de l'image en carré de 256x256"

        "on fait 24 images avec la taille de l'image "

        for x in range(0,6):

                for y in range(0,4):

                        Dictionnaire={}

                        imagcropC=imagecouleur[(7+y*256):(7+(y+1)*256),(4+x*256):(4+(x+1)*256)]

                        imagcropL=TotalUnique[(7+y*256):(7+(y+1)*256),(4+x*256):(4+(x+1)*256)]

                        labelunique=numpy.unique(imagcropL)

                        "sauvegarde du crop couleur et label associés"

                        io.imsave(namefile+"_"+str(nombre)+"_TU_C.tif",imagcropC,check_contrast=False)

                        io.imsave(namefile+"_"+str(nombre)+"_TU_L.tif",imagcropL,check_contrast=False)

                        for l in labelunique[1:]:

                                Dictionnaire[int(l)]=my_dict.get(l)

                        "sauvegarde du dico ave les labels indivs por les images 256*256"

                        tf1 = open(namefile+"_"+str(nombre)+"_TU.json", "w")

                        geojson.dump(Dictionnaire,tf1)

                        tf1.close()

                        nombre+=1



```



%% Cell type:code id: tags:



``` python

```

JSON_Qupath_to_ImageLabel.ipynb

0 → 100755
+110 −0
Changes for JSON_Qupath_to_ImageLabel.ipynb: 110 added lines, 0 removed lines.
Original line number Diff line number Diff line 
%% Cell type:code id: tags:



``` python

import geojson

from shapely.geometry import Polygon

from skimage.draw import polygon

import numpy as np

from skimage import io

import glob

```



%% Cell type:code id: tags:



``` python

def drawnucleiFinal(SizeImage,allshape):

    """

    SizeImage : size image

    allshape : coordinates of objects

    """

    TotalUnique=np.zeros((SizeImage[3],SizeImage[2]),np.uint16())



    compteurU=1

    Dictionnaire={}

    for objet in allshape:

        ROI=np.zeros((SizeImage[3],SizeImage[2]),np.uint8())

        if objet ["geometry"]["type"]=="Polygon":

            # Create a Polygon from the coordinates

            poly = Polygon(objet["geometry"]["coordinates"][0])



            "conversion of the polygon into a rounded numarray"

            contour=np.uint32(np.around(np.asarray(poly.exterior.coords)))



            "creation of the subtraction matrix to return to a start at 0,0"

            try:

                bc_arr, bc_row_means = np.broadcast_arrays(contour, np.array([SizeImage[0],SizeImage[1]]))

            except:

                print('erreur')

            polygone=bc_arr-bc_row_means

            "drawing the polygon on the image to make a mask"

            c,r=np.split(polygone,2,axis=1)



            rr, cc = polygon(r.reshape(r.shape[0]), c.reshape(c.shape[0]))

            tab=np.where(rr>=SizeImage[3])

            rr[tab]=SizeImage[3]-1



            tab=np.where(cc>=SizeImage[2])

            cc[tab]=SizeImage[2]-1



            try:

                ROI[rr,cc]=255

                Tab=np.where(ROI==255)

                if objet["properties"]["classification"]["name"]=="sombre":

                    Dictionnaire[compteurU]=1

                    TotalUnique[Tab]=compteurU

                    compteurU+=1



                if objet["properties"]["classification"]["name"]=="clair":

                    Dictionnaire[compteurU]=2

                    TotalUnique[Tab]=compteurU

                    compteurU+=1

            except :

                print("erreur 2")

                pass



    return TotalUnique,Dictionnaire

```



%% Cell type:code id: tags:



``` python

"folder containing outlines drawn on qupath and exported in json"

Liste=sorted(glob.glob("Examples/qupathProject/ExportJSON/*.json"))



"folder of images associated with contours"

dossierimage="Examples/"



"Read JSON files"

for json_detections in Liste:

        namefile=json_detections[:-5]

        #namefiletif=json_detections[:-5].split("/")[-1]+".tif"

        namefiletif=json_detections[:-19].split("/")[-1]

        with open(json_detections) as g:

                alldetections = geojson.load(g)





        "Contour design, labeling and dictionary creation for classification"

        TotalUnique,my_dict=drawnucleiFinal([0,0,1544,1038],alldetections)



        "read the original color image"

        imagecouleur=io.imread(dossierimage+namefiletif)

        nombre=0

        "Split of the image in 24 images of 256x256 pixels for the size of our current image "

        for x in range(0,6):

                for y in range(0,4):

                        Dictionnaire={}

                        imagcropC=imagecouleur[(7+y*256):(7+(y+1)*256),(4+x*256):(4+(x+1)*256)]

                        imagcropL=TotalUnique[(7+y*256):(7+(y+1)*256),(4+x*256):(4+(x+1)*256)]

                        labelunique=np.unique(imagcropL)

                        "saving the color crop image and associated label"

                        io.imsave(namefile+"_"+str(nombre)+"_TU_C.tif",imagcropC,check_contrast=False)

                        io.imsave(namefile+"_"+str(nombre)+"_TU_L.tif",imagcropL,check_contrast=False)

                        for l in labelunique[1:]:

                                Dictionnaire[int(l)]=my_dict.get(l)

                        "save dictionary with individual labels for images of size 256x256"

                        tf1 = open(namefile+"_"+str(nombre)+"_TU.json", "w")

                        geojson.dump(Dictionnaire,tf1)

                        tf1.close()

                        nombre+=1



```

Model_Build.ipynb

+5 −9
Changes for Model_Build.ipynb: 5 added lines, 9 removed lines.
Original line number Diff line number Diff line
@@ -742,9 +742,7 @@ 
  {

   "cell_type": "code",

   "execution_count": 24,

   "metadata": {

    "scrolled": false

   },

   "metadata": {},

   "outputs": [

    {

     "name": "stdout",
@@ -2034,9 +2032,7 @@ 
  {

   "cell_type": "code",

   "execution_count": 25,

   "metadata": {

    "scrolled": false

   },

   "metadata": {},

   "outputs": [

    {

     "name": "stderr",
@@ -2369,7 +2365,7 @@ 
   "hash": "31f2aee4e71d21fbe5cf8b01ff0e069b9275f58929596ceb00d14d90e3e16cd6"

  },

  "kernelspec": {

   "display_name": "Python 3",

   "display_name": "Python 3 (ipykernel)",

   "language": "python",

   "name": "python3"

  },
@@ -2383,9 +2379,9 @@ 
   "name": "python",

   "nbconvert_exporter": "python",

   "pygments_lexer": "ipython3",

   "version": "3.8.10"

   "version": "3.8.13"

  }

 },

 "nbformat": 4,

 "nbformat_minor": 2

 "nbformat_minor": 4

}