v1

.gitattributes

@@ -32,3 +32,6 @@ 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
+examples/*.* filter=lfs diff=lfs merge=lfs -text
+examples/*.* filter=lfs diff=lfs merge=lfs -text
+examples/*.* filter=lfs diff=lfs merge=lfs -text

app.py

@@ -0,0 +1,130 @@
+# Written by Dr Daniel Buscombe, Marda Science LLC
+#
+# MIT License
+#
+# Copyright (c) 2022, Marda Science LLC
+#
+# 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.
+
+import gradio as gr
+import os
+import pandas as pd
+import matplotlib.pyplot as plt
+# from skimage.transform import resize
+plt.style.use('fivethirtyeight')
+
+from funcs import *
+
+#================================================================
+def do_dgs(input_img):
+
+   maxscale=5 #20
+   verbose=True
+   x=0 #-0.5
+   resolution=1
+
+   w=input_img.shape[0]
+   h=input_img.shape[1]
+
+   # input_img = resize(input_img, (int(w/2), int(h/2)), preserve_range=True, clip=True) 
+   # resolution = 2
+
+   data_out = dgs(input_img, resolution, maxscale, verbose, x)
+
+   ## parse out dict into three separate dictionaries
+   stats = dict(list(data_out.items())[:4])
+   percentiles = dict(list(data_out.items())[4:6])
+   freqs_bins = dict(list(data_out.items())[6:])
+
+   if resolution!=1:
+      freqs_bins['grain size bins']*=resolution
+      percentiles['percentile_values']*=resolution
+
+      for k in stats.keys():
+            stats[k] = stats[k]*resolution
+
+   tmp = list(stats.keys())
+   d = {} 
+   for k in range(len(tmp)):
+       d.update( {tmp[k]: str(stats[tmp[k]])} )
+
+   pd.DataFrame(data=d.values(), index=d.keys()).to_csv('stats.csv')
+
+   tmp = list(percentiles.keys())
+   d2 = {} 
+   for k in range(len(tmp)):
+       d2.update( {tmp[k]: str(percentiles[tmp[k]])} )
+
+   pd.DataFrame(data=d2.values(), index=d2.keys()).to_csv('percentiles.csv')
+
+   # write each to csv file
+   pd.DataFrame.from_dict(freqs_bins).to_csv('freqs_bins.csv')
+
+   # plt.clf()
+   fig1=plt.figure(figsize=(4,4))
+   plt.plot(freqs_bins['grain size bins'], freqs_bins['grain size frequencies'], lw=2) 
+   plt.xlabel('Grain Size (pixels)')
+   plt.ylabel('Frequency')
+   plt.axvline(x=percentiles['percentile_values'][5], color="black", linestyle="--")
+   plt.text(percentiles['percentile_values'][5], .1, 'd50')
+   # plt.title('Grain Size Distribution')
+   # plt.xlim(np.maximum(0,stats['mean grain size']-(2*stats['grain size sorting'])), stats['mean grain size']+(2*stats['grain size sorting']))
+   # plt.savefig('psd.png', dpi=300, bbox_inches='tight')
+
+   # plt.clf()
+   fig2=plt.figure(figsize=(4,4))
+   plt.imshow(input_img,cmap='gray')
+   plt.plot([50, 50+stats['mean grain size']], [50, 50], 'r')
+   try:
+      plt.xlim(0,200)
+      plt.ylim(0,200)
+   except:
+      pass
+   plt.axis("off")
+
+   out = {}
+   out.update( {'mean grain size': float(str(stats['mean grain size'])[:4]) } )
+   out.update( {'grain size sorting': float(str(stats['grain size sorting'])[:4]) } )
+   out.update( {'d16': float(str(percentiles['percentile_values'][2])[:4]) } )
+   out.update( {'d50': float(str(percentiles['percentile_values'][5])[:4]) } )
+   out.update( {'d84': float(str(percentiles['percentile_values'][7])[:4]) } )
+   
+   return pd.DataFrame(data=out.values(), index=out.keys()).transpose(), fig1, fig2, 'stats.csv', 'percentiles.csv', 'freqs_bins.csv'
+
+ 
+  
+title = "Digital Grain Size" 
+description = "Upload an image of sediment. Download grain size stats. Clear between images. See https://github.com/dbuscombe-usgs/pyDGS. May take a very long time for large images" 
+    
+
+examples= [[l] for l in glob('examples/*.jpg')]
+
+inp = gr.Image(label='Upload images one-by-one')
+out1 = gr.Dataframe(label='Summary statistics', headers=["mean (px)", "sorting (px)", "d16 (px)", "d50 (px)", "d84 (px)"], type='pandas')
+#out1 = gr.outputs.Image(type='numpy')
+out2a = gr.Plot(label='Grain Size Distribution', type='auto') #'matplotlib')
+out2b = gr.Plot(label='Red Line = Mean Grain Length', type='auto') #'matplotlib')
+
+out4 = gr.File(label='Click to download grain size summary stats (px)')
+out5 = gr.File(label='Click to download grain size percentiles (px)')
+out6 = gr.File(label='Click to download grain size distribution: bins (px) and frequencies')
+
+Segapp = gr.Interface(do_dgs, inp, [out1, out2a, out2b, out4, out5, out6], title = title, description = description, examples=examples)
+
+Segapp.launch(enable_queue=True)

freqs_bins.csv

@@ -0,0 +1,171 @@
+,grain size frequencies,grain size bins
+0,0.0016426220390771682,1.8461538461538463
+1,0.002167585907438987,2.4615384615384617
+2,0.002668904649119935,3.0769230769230766
+3,0.0031699938478729063,3.6923076923076925
+4,0.003651220742984358,4.3076923076923075
+5,0.004126514308542332,4.923076923076923
+6,0.004609535210052074,5.538461538461538
+7,0.0050892227565306316,6.153846153846153
+8,0.0055744168004208915,6.769230769230769
+9,0.006045121494313346,7.384615384615385
+10,0.0065011214310461765,8.0
+11,0.006950538656204636,8.615384615384615
+12,0.007389333612356867,9.23076923076923
+13,0.007809496979602846,9.846153846153847
+14,0.008243336744138465,10.461538461538462
+15,0.008661944785145438,11.076923076923077
+16,0.009044087191385648,11.692307692307692
+17,0.00944831674873626,12.307692307692307
+18,0.009835436981769497,12.923076923076923
+19,0.010181128652331214,13.538461538461538
+20,0.01050415572102688,14.153846153846155
+21,0.010802845826556693,14.76923076923077
+22,0.011083779516659424,15.384615384615383
+23,0.011323672081431773,16.0
+24,0.01154374103241903,16.615384615384617
+25,0.01174549990392021,17.23076923076923
+26,0.011921139186435994,17.846153846153847
+27,0.012062611891841634,18.46153846153846
+28,0.012149605471212093,19.076923076923077
+29,0.012256719411607683,19.692307692307693
+30,0.012331703200822185,20.30769230769231
+31,0.012378469825517357,20.923076923076923
+32,0.012382949096099351,21.538461538461537
+33,0.012402253752891347,22.153846153846153
+34,0.012388518196972866,22.769230769230766
+35,0.012345729484965465,23.384615384615383
+36,0.012308219483070329,24.0
+37,0.012250705726838462,24.615384615384613
+38,0.012167377058765545,25.23076923076923
+39,0.012098696309146634,25.846153846153847
+40,0.011965261136264654,26.461538461538463
+41,0.011842219747470626,27.076923076923077
+42,0.011698744470455937,27.692307692307693
+43,0.011542259305387811,28.30769230769231
+44,0.011363379103230795,28.923076923076923
+45,0.011156752470320483,29.53846153846154
+46,0.010947912038893138,30.153846153846153
+47,0.010724109684386167,30.769230769230766
+48,0.010490633348516493,31.384615384615387
+49,0.010223843883055434,32.0
+50,0.00995192219751871,32.61538461538461
+51,0.00970494820986643,33.23076923076923
+52,0.009428785949242607,33.84615384615385
+53,0.009173145595160952,34.46153846153846
+54,0.008899741702773412,35.07692307692308
+55,0.008661039403923454,35.69230769230769
+56,0.008403869634455172,36.30769230769231
+57,0.008156783324095173,36.92307692307692
+58,0.0079269166346256,37.53846153846154
+59,0.007669992530057691,38.15384615384615
+60,0.007505031734891424,38.769230769230774
+61,0.007281079594547751,39.38461538461539
+62,0.007096595700886726,40.0
+63,0.006920991222684111,40.61538461538462
+64,0.006736086397618743,41.230769230769226
+65,0.006575650233596457,41.84615384615385
+66,0.0064181372103633404,42.46153846153846
+67,0.006274382395202138,43.07692307692307
+68,0.006141582712110023,43.69230769230769
+69,0.006008182294776169,44.30769230769231
+70,0.005885070587766238,44.92307692307693
+71,0.0057621280423813155,45.53846153846153
+72,0.005643066978396318,46.15384615384615
+73,0.005535288754723778,46.76923076923077
+74,0.005416295873642339,47.38461538461539
+75,0.005321391438623787,48.0
+76,0.005227047385243807,48.61538461538462
+77,0.005133002025558695,49.230769230769226
+78,0.005041083126273276,49.84615384615385
+79,0.004945032802174178,50.46153846153846
+80,0.004865689827376969,51.07692307692307
+81,0.004790519209525123,51.69230769230769
+82,0.00472369408648769,52.30769230769231
+83,0.004635627439684633,52.92307692307693
+84,0.004565541095126921,53.53846153846153
+85,0.004496305870199668,54.15384615384615
+86,0.004430789746276958,54.769230769230774
+87,0.004369170438953784,55.38461538461539
+88,0.004313165537727826,56.0
+89,0.004260803052558832,56.61538461538462
+90,0.004201367009515563,57.23076923076923
+91,0.004148478339295584,57.84615384615385
+92,0.004074701688990735,58.46153846153846
+93,0.0040520748074667975,59.07692307692308
+94,0.00400727588586632,59.69230769230769
+95,0.003956073673700605,60.30769230769231
+96,0.003913581667714583,60.92307692307692
+97,0.0038748269950508564,61.53846153846153
+98,0.003825921892423967,62.15384615384615
+99,0.0037973563463867908,62.769230769230774
+100,0.003745405560385463,63.38461538461539
+101,0.0037026294901179374,64.0
+102,0.00366627250047759,64.61538461538461
+103,0.0036119565983156955,65.23076923076923
+104,0.00360189208011268,65.84615384615384
+105,0.003559570569230386,66.46153846153847
+106,0.0035196429437868667,67.07692307692308
+107,0.0034877990085286346,67.6923076923077
+108,0.003457445037955179,68.3076923076923
+109,0.003424561985387016,68.92307692307692
+110,0.003392609732527566,69.53846153846153
+111,0.00336922764366803,70.15384615384616
+112,0.0033427922323019097,70.76923076923077
+113,0.0033212220795013608,71.38461538461539
+114,0.0032950670208449183,72.0
+115,0.00327952595609543,72.61538461538461
+116,0.0032654424561148598,73.23076923076923
+117,0.003242877258701247,73.84615384615384
+118,0.003232004290244402,74.46153846153845
+119,0.003220965181451411,75.07692307692308
+120,0.003212700078689291,75.6923076923077
+121,0.0032046498440294275,76.3076923076923
+122,0.0032006549619270803,76.92307692307692
+123,0.003216259993445007,77.53846153846155
+124,0.003211376135543289,78.15384615384616
+125,0.0031800941464485196,78.76923076923077
+126,0.0031904304207016933,79.38461538461539
+127,0.003196529413287816,80.0
+128,0.0032045771557983587,80.61538461538461
+129,0.003210219321174473,81.23076923076924
+130,0.0032157721848310597,81.84615384615385
+131,0.003223355288919453,82.46153846153845
+132,0.0032323883990270953,83.07692307692308
+133,0.0032426105888122185,83.6923076923077
+134,0.003250512700078592,84.3076923076923
+135,0.0032603325612916806,84.92307692307692
+136,0.0032680394631475263,85.53846153846153
+137,0.0032785397128673862,86.15384615384615
+138,0.003293319139721532,86.76923076923077
+139,0.0032984503116931093,87.38461538461539
+140,0.0033033505074822416,88.0
+141,0.003315195301571048,88.61538461538461
+142,0.0033200357101548974,89.23076923076923
+143,0.003328690924245227,89.84615384615385
+144,0.0033323933324966395,90.46153846153847
+145,0.003339397654829808,91.07692307692307
+146,0.0033409159213007664,91.6923076923077
+147,0.003347364526244621,92.3076923076923
+148,0.0033520913685139452,92.92307692307693
+149,0.0033514243713214757,93.53846153846153
+150,0.003352879546326893,94.15384615384615
+151,0.0033525735149703777,94.76923076923077
+152,0.003350254997413536,95.38461538461537
+153,0.003349566597363225,96.0
+154,0.0033452351633064693,96.61538461538463
+155,0.0033414811415823763,97.23076923076924
+156,0.003322441024870215,97.84615384615385
+157,0.003332490445771742,98.46153846153845
+158,0.0033265262827851957,99.07692307692308
+159,0.00331617847741188,99.6923076923077
+160,0.0033096442256116496,100.3076923076923
+161,0.0032994077570679646,100.92307692307692
+162,0.003289287964925445,101.53846153846155
+163,0.0032770381468205535,102.15384615384615
+164,0.003266991725648282,102.76923076923076
+165,0.0032576446351446145,103.38461538461539
+166,0.00324297091826369,104.0
+167,0.0032383175998767833,104.61538461538461
+168,0.0032074549401878236,105.23076923076924
+169,0.0032004106205451976,105.84615384615385

funcs.py

@@ -0,0 +1,213 @@
+# Written by Dr Daniel Buscombe, Marda Science LLC
+#
+# MIT License
+#
+# Copyright (c) 2022, Marda Science LLC
+#
+# 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.
+
+# from imageio import imread
+import pywt
+#from tqdm import tqdm
+from skimage.restoration import denoise_wavelet, estimate_sigma
+from functools import partial
+# rescale_sigma=True required to silence deprecation warnings
+_denoise_wavelet = partial(denoise_wavelet, rescale_sigma=True)
+import numpy as np
+import scipy.stats as stats
+from glob import glob
+
+def rescale(dat,mn,mx):
+    """
+    rescales an input dat between mn and mx
+    """
+    m = min(dat.flatten())
+    M = max(dat.flatten())
+    return (mx-mn)*(dat-m)/(M-m)+mn
+
+##====================================
+def standardize(img):
+    img = np.array(img)
+    #standardization using adjusted standard deviation
+    N = np.shape(img)[0] * np.shape(img)[1]
+    s = np.maximum(np.std(img), 1.0/np.sqrt(N))
+    m = np.mean(img)
+    img = (img - m) / s
+    img = rescale(img, 0, 1)
+    del m, s, N
+
+    return img
+
+
+# =========================================================
+# =========================================================
+def dgs(input_img, resolution=1, maxscale=4, verbose=1, x=-0.5):
+
+   #if verbose==1:
+   print("===========================================")
+   print("======DIGITAL GRAIN SIZE: WAVELET==========")
+   print("===========================================")
+   print("=CALCULATE GRAIN SIZE-DISTRIBUTION FROM AN=")
+   print("====IMAGE OF SEDIMENT/GRANULAR MATERIAL====")
+   print("===========================================")
+   print("======A PROGRAM BY DANIEL BUSCOMBE=========")
+   print("====MARDASCIENCE, FLAGSTAFF, ARIZONA=======")
+   print("========REVISION 4.2, APR 2022===========")
+   print("===========================================")
+
+   # ======= stage 1 ==========================
+   #read image
+   if verbose==1:
+      print("~~~~~~~~~~~~~~~~~~~~~~~~~~~~")
+      print('Processing image ')
+
+   im = np.squeeze(input_img)  # squeeze singleton dimensions
+   if len(np.shape(im))>3:
+       im = im[:, :, :3]            # only keep the first 3 bands
+
+   if len(np.shape(im))==3: # if rgb, convert to grey
+      im = (0.299 * im[:,:,0] + 0.5870*im[:,:,1] + 0.114*im[:,:,2]).astype('uint8')
+
+   nx,ny = np.shape(im)
+   if nx>ny:
+      im=im.T
+
+   im = standardize(im)
+   # # ======= stage 2 ==========================
+   # Denoised image using default parameters of `denoise_wavelet`
+   filter=False
+
+   if filter:
+
+      sigma_est = estimate_sigma(im, multichannel=False, average_sigmas=True)
+      region = denoise_wavelet(im, multichannel=False, rescale_sigma=True,
+                                 method='VisuShrink', mode='soft', sigma=sigma_est)
+
+   else:
+
+      region = im.copy()
+
+   original = rescale(region,0,255)
+
+   nx, ny = original.shape
+
+   # ======= stage 3 ==========================
+   # call cwt to get particle size distribution
+
+
+   ## initial guess
+   P = []; M = []
+   for k in np.linspace(1,nx-1,40):
+      [cfs, frequencies] = pywt.cwt(original[int(k),:], np.arange(3, np.maximum(nx,ny)/maxscale, 1),  'morl' , .5)
+      period = 1. / frequencies
+      power =(abs(cfs)) ** 2
+      power = np.mean(np.abs(power), axis=1)/(period**2)
+      P.append(power)
+
+      M.append(period[np.argmax(power)])
+
+   p = np.mean(np.vstack(P), axis=0)
+   p = np.array(p/np.sum(p))
+
+   # get real scales by multiplying by resolution (mm/pixel)
+   scales = np.array(period)*resolution
+
+   print(np.sum(p*scales))
+   if np.sum(p*scales)>80:
+      x=1
+      maxscale=4
+   elif (np.sum(p*scales)<80) and (np.sum(p*scales)>60):
+      x=0.75
+      maxscale=8
+   elif (np.sum(p*scales)<60) and (np.sum(p*scales)>40):
+      x=0.5
+      maxscale=12
+   elif (np.sum(p*scales)<40) and (np.sum(p*scales)>20):
+      x=-0.5
+      maxscale=16
+   elif np.sum(p*scales)<20:
+      x=-1
+      maxscale=20
+
+   print("x is {}".format(x))  
+   print("maxscale is {}".format(maxscale))  
+
+   ## for real
+   P = []; M = []
+   for k in np.linspace(1,nx-1,100):
+      [cfs, frequencies] = pywt.cwt(original[int(k),:], np.arange(3, np.maximum(nx,ny)/maxscale, 1),  'morl' , .5)
+      period = 1. / frequencies
+      power =(abs(cfs)) ** 2
+      power = np.mean(np.abs(power), axis=1)/(period**2)
+      P.append(power)
+
+      M.append(period[np.argmax(power)])
+
+   p = np.mean(np.vstack(P), axis=0)
+   p = np.array(p/np.sum(p))
+
+   # get real scales by multiplying by resolution (mm/pixel)
+   scales = np.array(period)*resolution
+
+   srt = np.sqrt(np.sum(p*((scales-np.mean(M))**2)))
+
+   # plt.plot(scales, p,'m', lw=2)
+
+   p = p+stats.norm.pdf(scales, np.mean(M), srt/np.pi)
+   p = p/np.sum(p)
+
+   mnsz = np.sum(p*scales)
+   srt = np.sqrt(np.sum(p*((scales-mnsz)**2)))
+
+   ind =np.where(scales < (mnsz+3*srt))[0]
+   scales= scales[ind]
+   p = p[ind]
+   # p = np.hstack([0,p])
+   # scales = np.hstack([0,scales])   
+
+   # area-by-number to volume-by-number
+   r_v = (p*scales**x) / np.sum(p*scales**x) #volume-by-weight proportion
+
+   # ======= stage 5 ==========================
+   # calc particle size stats
+
+   pd = np.interp([.05,.1,.16,.25,.3,.5,.75,.84,.9,.95],np.hstack((0,np.cumsum(r_v))), np.hstack((0,scales)) )
+   if verbose==1:
+      print("d50 = "+str(pd[4]))
+
+   mnsz = np.sum(r_v*scales)
+   if verbose==1:
+      print("mean size = "+str(mnsz))
+
+   srt = np.sqrt(np.sum(r_v*((scales-mnsz)**2)))
+   if verbose==1:
+      print("stdev = "+str(srt))
+
+   sk = (sum(r_v*((scales-mnsz)**3)))/(100*srt**3)
+   if verbose==1:
+      print("skewness = "+str(sk))
+
+   kurt = (sum(r_v*((scales-mnsz)**4)))/(100*srt**4)
+   if verbose==1:
+      print("kurtosis = "+str(kurt))
+
+   # ======= stage 6 ==========================
+   # return a dict object of stats
+   return {'mean grain size': mnsz, 'grain size sorting': srt, 'grain size skewness': sk, 'grain size kurtosis': kurt, 'percentiles': [.05,.1,.16,.25,.3,.5,.75,.84,.9,.95], 'percentile_values': pd, 'grain size frequencies': r_v, 'grain size bins': scales}
+

percentiles.csv

@@ -0,0 +1,4 @@
+,0
+percentiles,"[0.05, 0.1, 0.16, 0.25, 0.3, 0.5, 0.75, 0.84, 0.9, 0.95]"
+percentile_values,"[ 8.42088522 12.12575912 15.64543534 20.28097679 22.76488214 33.69610315
+ 60.00452672 75.82437868 87.2558193  96.47342121]"

requirements.txt

@@ -0,0 +1,6 @@
+numpy
+matplotlib
+imageio
+scikit-image
+pywavelets
+pandas

stats.csv

@@ -0,0 +1,5 @@
+,0
+mean grain size,42.10146327816689
+grain size sorting,27.384385416408026
+grain size skewness,0.007339603505687132
+grain size kurtosis,0.024435709503563517