examples/N2V-Convallaria-1-Training.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# N2V Training\n",
    "Here we will train a plain N2V network on single noisy images."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "CUDA available? True\n"
     ]
    }
   ],
   "source": [
    "import matplotlib.pyplot as plt\n",
    "from pn2v.unet import UNet\n",
    "\n",
    "from pn2v import utils\n",
    "from pn2v import training\n",
    "from tifffile import imread\n",
    "\n",
    "# See if we can use a GPU\n",
    "device=utils.getDevice()"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Load Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "path='data/Convallaria_diaphragm/'\n",
    "\n",
    "# Load the training data\n",
    "data=imread(path+'20190520_tl_25um_50msec_05pc_488_130EM_Conv.tif')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Create the Network and Train it\n",
    "This can take a while."
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "/home/krull/projects/pn2vCheck/pn2v/unet/model.py:215: UserWarning: nn.init.xavier_normal is now deprecated in favor of nn.init.xavier_normal_.\n",
      "  init.xavier_normal(m.weight)\n",
      "/home/krull/projects/pn2vCheck/pn2v/unet/model.py:216: UserWarning: nn.init.constant is now deprecated in favor of nn.init.constant_.\n",
      "  init.constant(m.bias, 0)\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 0 finished      \n",
      "avg. loss: 0.3765426553785801+-(2SEM)0.1279227916685506      \n",
      "Epoch 1 finished      \n",
      "avg. loss: 0.2777821462601423+-(2SEM)0.09849123862890843      \n",
      "Epoch 2 finished      \n",
      "avg. loss: 0.13786188121885062+-(2SEM)0.042484115623708164      \n",
      "Epoch 3 finished      \n",
      "avg. loss: 0.1334076628088951+-(2SEM)0.030446528995180064      \n",
      "Epoch 4 finished      \n",
      "avg. loss: 0.1570326631888747+-(2SEM)0.06266355305142955      \n",
      "Epoch 5 finished      \n",
      "avg. loss: 0.11786636784672737+-(2SEM)0.024805765654469018      \n",
      "Epoch 6 finished      \n",
      "avg. loss: 0.12015673499554395+-(2SEM)0.038688107814813404      \n",
      "Epoch 7 finished      \n",
      "avg. loss: 0.15155412238091232+-(2SEM)0.03580783946044995      \n",
      "Epoch 8 finished      \n",
      "avg. loss: 0.13173507079482077+-(2SEM)0.03550308823546951      \n",
      "Epoch 9 finished      \n",
      "avg. loss: 0.10643239542841912+-(2SEM)0.025437866535106372      \n",
      "Epoch 10 finished      \n",
      "avg. loss: 0.1390177359804511+-(2SEM)0.03262271473702893      \n",
      "Epoch 11 finished      \n",
      "avg. loss: 0.10688655823469162+-(2SEM)0.03386666999237347      \n",
      "Epoch 12 finished      \n",
      "avg. loss: 0.08581975689157843+-(2SEM)0.01800878489833051      \n",
      "Epoch 13 finished      \n",
      "avg. loss: 0.11406578775495291+-(2SEM)0.023251838583189452      \n",
      "Epoch 14 finished      \n",
      "avg. loss: 0.14059206303209065+-(2SEM)0.03290585528038259      \n",
      "Epoch 15 finished      \n",
      "avg. loss: 0.09359127879142762+-(2SEM)0.0241119450348921      \n",
      "Epoch 16 finished      \n",
      "avg. loss: 0.12195333242416381+-(2SEM)0.02501105516340815      \n",
      "Epoch 17 finished      \n",
      "avg. loss: 0.12988977916538716+-(2SEM)0.03815226688464073      \n",
      "Epoch 18 finished      \n",
      "avg. loss: 0.12366882730275393+-(2SEM)0.027189125125003336      \n",
      "Epoch 19 finished      \n",
      "avg. loss: 0.10210984116420149+-(2SEM)0.022418380689858776      \n",
      "Epoch 20 finished      \n",
      "avg. loss: 0.12444097828119993+-(2SEM)0.029045603225091622      \n",
      "Epoch 21 finished      \n",
      "avg. loss: 0.10741706043481827+-(2SEM)0.02822903879313525      \n",
      "Epoch 22 finished      \n",
      "avg. loss: 0.09048820296302437+-(2SEM)0.022297301252983473      \n",
      "Epoch 23 finished      \n",
      "avg. loss: 0.1712287465110421+-(2SEM)0.045013099400308405      \n",
      "Epoch 24 finished      \n",
      "avg. loss: 0.11313214357942343+-(2SEM)0.022961074805972086      \n",
      "Epoch 25 finished      \n",
      "avg. loss: 0.1349453882314265+-(2SEM)0.03778377555736207      \n",
      "Epoch 26 finished      \n",
      "avg. loss: 0.12360266987234354+-(2SEM)0.035055524429416916      \n",
      "Epoch 27 finished      \n",
      "avg. loss: 0.13626814894378186+-(2SEM)0.04142801321430307      \n",
      "Epoch 28 finished      \n",
      "avg. loss: 0.11804973399266601+-(2SEM)0.038445151843098484      \n",
      "Epoch 29 finished      \n",
      "avg. loss: 0.12746565323323011+-(2SEM)0.03204152732292345      \n",
      "Epoch 30 finished      \n",
      "avg. loss: 0.10145141184329987+-(2SEM)0.023160364589701603      \n",
      "Epoch 31 finished      \n",
      "avg. loss: 0.10493494812399148+-(2SEM)0.019925721409395577      \n",
      "Epoch 32 finished      \n",
      "avg. loss: 0.1020697253756225+-(2SEM)0.035299514426542666      \n",
      "Epoch 33 finished      \n",
      "avg. loss: 0.12627474814653397+-(2SEM)0.03804131974844131      \n",
      "Epoch 34 finished      \n",
      "avg. loss: 0.1115457646548748+-(2SEM)0.030952622836915213      \n",
      "Epoch 35 finished      \n",
      "avg. loss: 0.09749468695372343+-(2SEM)0.019155353528757357      \n",
      "Epoch 36 finished      \n",
      "avg. loss: 0.09871908966451884+-(2SEM)0.027268469009675462      \n",
      "Epoch 37 finished      \n",
      "avg. loss: 0.13428890481591224+-(2SEM)0.03173196034249195      \n",
      "Epoch 38 finished      \n",
      "avg. loss: 0.10146755557507277+-(2SEM)0.024181747695418993      \n",
      "Epoch 39 finished      \n",
      "avg. loss: 0.1500510662794113+-(2SEM)0.04375110361980511      \n",
      "Epoch 40 finished      \n",
      "avg. loss: 0.10287422835826873+-(2SEM)0.03230799453536254      \n",
      "Epoch 41 finished      \n",
      "avg. loss: 0.13052849434316158+-(2SEM)0.03887099680368908      \n",
      "Epoch 42 finished      \n",
      "avg. loss: 0.09415352856740355+-(2SEM)0.021721775663152483      \n",
      "Epoch 43 finished      \n",
      "avg. loss: 0.0934420321136713+-(2SEM)0.020704164425360452      \n",
      "Epoch 44 finished      \n",
      "avg. loss: 0.12231855383142828+-(2SEM)0.0373140153669821      \n",
      "Epoch    44: reducing learning rate of group 0 to 5.0000e-04.\n",
      "Epoch 45 finished      \n",
      "avg. loss: 0.12197865787893533+-(2SEM)0.0350774280279961      \n",
      "Epoch 46 finished      \n",
      "avg. loss: 0.12005211114883423+-(2SEM)0.03810112999342088      \n",
      "Epoch 47 finished      \n",
      "avg. loss: 0.12696708776056767+-(2SEM)0.02945083989003774      \n",
      "Epoch 48 finished      \n",
      "avg. loss: 0.11856234353035688+-(2SEM)0.03904329873142306      \n",
      "Epoch 49 finished      \n",
      "avg. loss: 0.10823339000344276+-(2SEM)0.03314748155815287      \n",
      "Epoch 50 finished      \n",
      "avg. loss: 0.09282249361276626+-(2SEM)0.018483206802924996      \n",
      "Epoch 51 finished      \n",
      "avg. loss: 0.10986562203615904+-(2SEM)0.026196323114124023      \n",
      "Epoch 52 finished      \n",
      "avg. loss: 0.11989624183624983+-(2SEM)0.03335820736669722      \n",
      "Epoch 53 finished      \n",
      "avg. loss: 0.10963170025497675+-(2SEM)0.033920954139544934      \n",
      "Epoch 54 finished      \n",
      "avg. loss: 0.10676916232332587+-(2SEM)0.028780547867914483      \n",
      "Epoch 55 finished      \n",
      "avg. loss: 0.09472140846773983+-(2SEM)0.01993356731753518      \n",
      "Epoch    55: reducing learning rate of group 0 to 2.5000e-04.\n",
      "Epoch 56 finished      \n",
      "avg. loss: 0.14791604038327932+-(2SEM)0.03778694915830772      \n",
      "Epoch 57 finished      \n",
      "avg. loss: 0.11597967119887471+-(2SEM)0.02010362578760813      \n",
      "Epoch 58 finished      \n",
      "avg. loss: 0.1096066426485777+-(2SEM)0.030767395431136074      \n",
      "Epoch 59 finished      \n",
      "avg. loss: 0.1016509854234755+-(2SEM)0.029667027485687923      \n",
      "Epoch 60 finished      \n",
      "avg. loss: 0.1270496379584074+-(2SEM)0.03495777228329363      \n",
      "Epoch 61 finished      \n",
      "avg. loss: 0.09260655343532562+-(2SEM)0.024576434895243856      \n",
      "Epoch 62 finished      \n",
      "avg. loss: 0.11635740473866463+-(2SEM)0.03346637376119446      \n",
      "Epoch 63 finished      \n",
      "avg. loss: 0.12848153430968523+-(2SEM)0.026134312930378864      \n",
      "Epoch 64 finished      \n",
      "avg. loss: 0.09276588819921017+-(2SEM)0.016391741674188876      \n",
      "Epoch 65 finished      \n",
      "avg. loss: 0.09451652597635984+-(2SEM)0.017604726525840268      \n",
      "Epoch 66 finished      \n",
      "avg. loss: 0.08814101815223693+-(2SEM)0.01977715110641835      \n",
      "Epoch    66: reducing learning rate of group 0 to 1.2500e-04.\n",
      "Epoch 67 finished      \n",
      "avg. loss: 0.09097991231828928+-(2SEM)0.022883339629358835      \n",
      "Epoch 68 finished      \n",
      "avg. loss: 0.09329799227416516+-(2SEM)0.019301932803493523      \n",
      "Epoch 69 finished      \n",
      "avg. loss: 0.11130706444382668+-(2SEM)0.021986366726675127      \n",
      "Epoch 70 finished      \n",
      "avg. loss: 0.11211018385365605+-(2SEM)0.03438991716637538      \n",
      "Epoch 71 finished      \n",
      "avg. loss: 0.10199320409446955+-(2SEM)0.02100097271253441      \n",
      "Epoch 72 finished      \n",
      "avg. loss: 0.11935705058276654+-(2SEM)0.03346069762459516      \n",
      "Epoch 73 finished      \n",
      "avg. loss: 0.1103150738403201+-(2SEM)0.02748598569169171      \n",
      "Epoch 74 finished      \n",
      "avg. loss: 0.08751724380999804+-(2SEM)0.014104603009597665      \n",
      "Epoch 75 finished      \n",
      "avg. loss: 0.10922938166186213+-(2SEM)0.03337654010679488      \n",
      "Epoch 76 finished      \n",
      "avg. loss: 0.11687524262815714+-(2SEM)0.033520844639186045      \n",
      "Epoch 77 finished      \n",
      "avg. loss: 0.08763003554195166+-(2SEM)0.01679892352066965      \n",
      "Epoch 78 finished      \n",
      "avg. loss: 0.11434677932411433+-(2SEM)0.03147553083207084      \n",
      "Epoch    78: reducing learning rate of group 0 to 6.2500e-05.\n",
      "Epoch 79 finished      \n",
      "avg. loss: 0.10752540864050389+-(2SEM)0.01979842648909614      \n",
      "Epoch 80 finished      \n",
      "avg. loss: 0.12502142060548066+-(2SEM)0.028414932618222574      \n",
      "Epoch 81 finished      \n",
      "avg. loss: 0.11060827933251857+-(2SEM)0.025346578570097006      \n",
      "Epoch 82 finished      \n",
      "avg. loss: 0.10329240560531616+-(2SEM)0.021232690045960175      \n",
      "Epoch 83 finished      \n",
      "avg. loss: 0.11287378612905741+-(2SEM)0.03807167175691467      \n",
      "Epoch 84 finished      \n",
      "avg. loss: 0.11093791034072638+-(2SEM)0.028182385282816476      \n",
      "Epoch 85 finished      \n",
      "avg. loss: 0.11638402361422777+-(2SEM)0.030890167152898603      \n",
      "Epoch 86 finished      \n",
      "avg. loss: 0.12947785584256052+-(2SEM)0.03196813853179929      \n",
      "Epoch 87 finished      \n",
      "avg. loss: 0.10338967423886061+-(2SEM)0.02798442211400995      \n",
      "Epoch 88 finished      \n",
      "avg. loss: 0.10320335691794753+-(2SEM)0.03536426086019903      \n",
      "Epoch 89 finished      \n",
      "avg. loss: 0.11111485827714204+-(2SEM)0.024862690681056787      \n",
      "Epoch    89: reducing learning rate of group 0 to 3.1250e-05.\n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 90 finished      \n",
      "avg. loss: 0.12674069879576563+-(2SEM)0.026610592299468022      \n",
      "Epoch 91 finished      \n",
      "avg. loss: 0.11817380674183368+-(2SEM)0.042820805172960275      \n",
      "Epoch 92 finished      \n",
      "avg. loss: 0.14416311774402857+-(2SEM)0.038800917523801125      \n",
      "Epoch 93 finished      \n",
      "avg. loss: 0.09693180341273547+-(2SEM)0.024967627046776256      \n",
      "Epoch 94 finished      \n",
      "avg. loss: 0.08443844094872474+-(2SEM)0.017988774552256667      \n",
      "Epoch 95 finished      \n",
      "avg. loss: 0.11432975083589554+-(2SEM)0.024570183745368387      \n",
      "Epoch 96 finished      \n",
      "avg. loss: 0.09527028426527977+-(2SEM)0.01981979311988979      \n",
      "Epoch 97 finished      \n",
      "avg. loss: 0.1184925596229732+-(2SEM)0.03931482718061687      \n",
      "Epoch 98 finished      \n",
      "avg. loss: 0.1318995039910078+-(2SEM)0.030175598140719632      \n",
      "Epoch 99 finished      \n",
      "avg. loss: 0.13054412174969912+-(2SEM)0.03888581029423797      \n",
      "Epoch 100 finished      \n",
      "avg. loss: 0.1154963007196784+-(2SEM)0.025382615854024404      \n",
      "Epoch   100: reducing learning rate of group 0 to 1.5625e-05.\n",
      "Epoch 101 finished      \n",
      "avg. loss: 0.11740099526941776+-(2SEM)0.03186208771543293      \n",
      "Epoch 102 finished      \n",
      "avg. loss: 0.11018431875854731+-(2SEM)0.0235748742018259      \n",
      "Epoch 103 finished      \n",
      "avg. loss: 0.0978842101059854+-(2SEM)0.017965261152470848      \n",
      "Epoch 104 finished      \n",
      "avg. loss: 0.11955860331654548+-(2SEM)0.02828509620277827      \n",
      "Epoch 105 finished      \n",
      "avg. loss: 0.11859250236302614+-(2SEM)0.032301274103039505      \n",
      "Epoch 106 finished      \n",
      "avg. loss: 0.1250128386542201+-(2SEM)0.034482520082020324      \n",
      "Epoch 107 finished      \n",
      "avg. loss: 0.1171593495644629+-(2SEM)0.03694265871551958      \n",
      "Epoch 108 finished      \n",
      "avg. loss: 0.1261391805484891+-(2SEM)0.030693555428506415      \n",
      "Epoch 109 finished      \n",
      "avg. loss: 0.12123149670660496+-(2SEM)0.03287548459505225      \n",
      "Epoch 110 finished      \n",
      "avg. loss: 0.1384928349405527+-(2SEM)0.041656524120669114      \n",
      "Epoch 111 finished      \n",
      "avg. loss: 0.096132099814713+-(2SEM)0.019026419630531696      \n",
      "Epoch   111: reducing learning rate of group 0 to 7.8125e-06.\n",
      "Epoch 112 finished      \n",
      "avg. loss: 0.13461431320756673+-(2SEM)0.03669008101430109      \n",
      "Epoch 113 finished      \n",
      "avg. loss: 0.1316376967355609+-(2SEM)0.038492012526563764      \n",
      "Epoch 114 finished      \n",
      "avg. loss: 0.12863745223730802+-(2SEM)0.025594713630798523      \n",
      "Epoch 115 finished      \n",
      "avg. loss: 0.10547595396637917+-(2SEM)0.028306350067394286      \n",
      "Epoch 116 finished      \n",
      "avg. loss: 0.11494108960032463+-(2SEM)0.026761310710394545      \n",
      "Epoch 117 finished      \n",
      "avg. loss: 0.10668933186680078+-(2SEM)0.024808740228443194      \n",
      "Epoch 118 finished      \n",
      "avg. loss: 0.10839723143726587+-(2SEM)0.02980732334223127      \n",
      "Epoch 119 finished      \n",
      "avg. loss: 0.11011406732723117+-(2SEM)0.02562807646937819      \n",
      "Epoch 120 finished      \n",
      "avg. loss: 0.09530323948711157+-(2SEM)0.018447752469698993      \n",
      "Epoch 121 finished      \n",
      "avg. loss: 0.10099674947559834+-(2SEM)0.021526935801222995      \n",
      "Epoch 122 finished      \n",
      "avg. loss: 0.09926179945468902+-(2SEM)0.017567999491101667      \n",
      "Epoch   122: reducing learning rate of group 0 to 3.9063e-06.\n",
      "Epoch 123 finished      \n",
      "avg. loss: 0.11076054070144892+-(2SEM)0.022261528896833242      \n",
      "Epoch 124 finished      \n",
      "avg. loss: 0.09971698634326458+-(2SEM)0.018123436627957106      \n",
      "Epoch 125 finished      \n",
      "avg. loss: 0.09761321097612381+-(2SEM)0.023717244156330367      \n",
      "Epoch 126 finished      \n",
      "avg. loss: 0.09561513289809227+-(2SEM)0.02301536480224505      \n",
      "Epoch 127 finished      \n",
      "avg. loss: 0.10759383253753185+-(2SEM)0.030026559176441873      \n",
      "Epoch 128 finished      \n",
      "avg. loss: 0.10621486064046622+-(2SEM)0.020626933856791943      \n",
      "Epoch 129 finished      \n",
      "avg. loss: 0.13129500653594733+-(2SEM)0.04097389321090376      \n",
      "Epoch 130 finished      \n",
      "avg. loss: 0.11718312669545412+-(2SEM)0.03106611041421776      \n",
      "Epoch 131 finished      \n",
      "avg. loss: 0.11586161898449063+-(2SEM)0.03206679436805331      \n",
      "Epoch 132 finished      \n",
      "avg. loss: 0.11611682623624801+-(2SEM)0.0343968297364438      \n",
      "Epoch 133 finished      \n",
      "avg. loss: 0.11722639109939337+-(2SEM)0.02621681765428275      \n",
      "Epoch   133: reducing learning rate of group 0 to 1.9531e-06.\n",
      "Epoch 134 finished      \n",
      "avg. loss: 0.09868948049843311+-(2SEM)0.027094548168069287      \n",
      "Epoch 135 finished      \n",
      "avg. loss: 0.10981661286205054+-(2SEM)0.026967147130728154      \n",
      "Epoch 136 finished      \n",
      "avg. loss: 0.13558879420161246+-(2SEM)0.03558482371558996      \n",
      "Epoch 137 finished      \n",
      "avg. loss: 0.11569489520043134+-(2SEM)0.025578560271005454      \n",
      "Epoch 138 finished      \n",
      "avg. loss: 0.135537800937891+-(2SEM)0.029066546838919297      \n",
      "Epoch 139 finished      \n",
      "avg. loss: 0.11381082721054554+-(2SEM)0.03585940211117464      \n",
      "Epoch 140 finished      \n",
      "avg. loss: 0.1061279846355319+-(2SEM)0.02322806330903561      \n",
      "Epoch 141 finished      \n",
      "avg. loss: 0.08050485998392105+-(2SEM)0.013756020951108186      \n",
      "Epoch 142 finished      \n",
      "avg. loss: 0.10240876656025648+-(2SEM)0.021192030970746966      \n",
      "Epoch 143 finished      \n",
      "avg. loss: 0.10718167666345835+-(2SEM)0.029581080250312627      \n",
      "Epoch 144 finished      \n",
      "avg. loss: 0.08856086386367679+-(2SEM)0.019279726461596047      \n",
      "Epoch   144: reducing learning rate of group 0 to 9.7656e-07.\n",
      "Epoch 145 finished      \n",
      "avg. loss: 0.10055750207975507+-(2SEM)0.027928956892122964      \n",
      "Epoch 146 finished      \n",
      "avg. loss: 0.11833430789411067+-(2SEM)0.030901492327499495      \n",
      "Epoch 147 finished      \n",
      "avg. loss: 0.13187403623014687+-(2SEM)0.03545853187736862      \n",
      "Epoch 148 finished      \n",
      "avg. loss: 0.10685692559927702+-(2SEM)0.03735066745827625      \n",
      "Epoch 149 finished      \n",
      "avg. loss: 0.13163984306156634+-(2SEM)0.031293379152613654      \n",
      "Epoch 150 finished      \n",
      "avg. loss: 0.0866280758753419+-(2SEM)0.025490155435690432      \n",
      "Epoch 151 finished      \n",
      "avg. loss: 0.10372214205563068+-(2SEM)0.03079476630457483      \n",
      "Epoch 152 finished      \n",
      "avg. loss: 0.08386387396603823+-(2SEM)0.01785308603517603      \n",
      "Epoch 153 finished      \n",
      "avg. loss: 0.137422432936728+-(2SEM)0.03727949326564278      \n",
      "Epoch 154 finished      \n",
      "avg. loss: 0.13285119831562042+-(2SEM)0.036760325655467746      \n",
      "Epoch 155 finished      \n",
      "avg. loss: 0.11460651978850364+-(2SEM)0.02565690460466736      \n",
      "Epoch   155: reducing learning rate of group 0 to 4.8828e-07.\n",
      "Epoch 156 finished      \n",
      "avg. loss: 0.10607877243310213+-(2SEM)0.02610936927372234      \n",
      "Epoch 157 finished      \n",
      "avg. loss: 0.10543674966320396+-(2SEM)0.030091650946047977      \n",
      "Epoch 158 finished      \n",
      "avg. loss: 0.1317719966173172+-(2SEM)0.026103852702800332      \n",
      "Epoch 159 finished      \n",
      "avg. loss: 0.12992613650858403+-(2SEM)0.03045644829020848      \n",
      "Epoch 160 finished      \n",
      "avg. loss: 0.12119826711714268+-(2SEM)0.03812833172816358      \n",
      "Epoch 161 finished      \n",
      "avg. loss: 0.11968649122864009+-(2SEM)0.030948653182784946      \n",
      "Epoch 162 finished      \n",
      "avg. loss: 0.12116792034357786+-(2SEM)0.0348117954080701      \n",
      "Epoch 163 finished      \n",
      "avg. loss: 0.10664117094129325+-(2SEM)0.02839936570502564      \n",
      "Epoch 164 finished      \n",
      "avg. loss: 0.14187282845377922+-(2SEM)0.033132942656715854      \n",
      "Epoch 165 finished      \n",
      "avg. loss: 0.10885346904397011+-(2SEM)0.021053670279203716      \n",
      "Epoch 166 finished      \n",
      "avg. loss: 0.10041732573881745+-(2SEM)0.024639049569053745      \n",
      "Epoch   166: reducing learning rate of group 0 to 2.4414e-07.\n",
      "Epoch 167 finished      \n",
      "avg. loss: 0.1326250931248069+-(2SEM)0.03460988616153599      \n",
      "Epoch 168 finished      \n",
      "avg. loss: 0.11015231441706419+-(2SEM)0.031406496293688746      \n",
      "Epoch 169 finished      \n",
      "avg. loss: 0.10257368050515651+-(2SEM)0.02149557700267231      \n",
      "Epoch 170 finished      \n",
      "avg. loss: 0.12168299518525601+-(2SEM)0.02888712208416637      \n",
      "Epoch 171 finished      \n",
      "avg. loss: 0.09749463051557541+-(2SEM)0.01970208311135665      \n",
      "Epoch 172 finished      \n",
      "avg. loss: 0.08430260717868805+-(2SEM)0.01729612687672161      \n",
      "Epoch 173 finished      \n",
      "avg. loss: 0.0909187407232821+-(2SEM)0.017904041119353155      \n",
      "Epoch 174 finished      \n",
      "avg. loss: 0.13149883840233087+-(2SEM)0.04163274308045978      \n",
      "Epoch 175 finished      \n",
      "avg. loss: 0.10299822539091111+-(2SEM)0.01959118271310663      \n",
      "Epoch 176 finished      \n",
      "avg. loss: 0.10632069148123265+-(2SEM)0.02441856018318787      \n",
      "Epoch 177 finished      \n"
     ]
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "avg. loss: 0.10577574456110597+-(2SEM)0.030708221441462142      \n",
      "Epoch   177: reducing learning rate of group 0 to 1.2207e-07.\n",
      "Epoch 178 finished      \n",
      "avg. loss: 0.09064709218218922+-(2SEM)0.0238484442746126      \n",
      "Epoch 179 finished      \n",
      "avg. loss: 0.11593986582010984+-(2SEM)0.028848489746458355      \n",
      "Epoch 180 finished      \n",
      "avg. loss: 0.11242860294878483+-(2SEM)0.033348526154819605      \n",
      "Epoch 181 finished      \n",
      "avg. loss: 0.12201768085360527+-(2SEM)0.022947279218883468      \n",
      "Epoch 182 finished      \n",
      "avg. loss: 0.10489928144961595+-(2SEM)0.019302167214129973      \n",
      "Epoch 183 finished      \n",
      "avg. loss: 0.12149366810917854+-(2SEM)0.03247195558397946      \n",
      "Epoch 184 finished      \n",
      "avg. loss: 0.12925710380077363+-(2SEM)0.028510144869626143      \n",
      "Epoch 185 finished      \n",
      "avg. loss: 0.11504415702074766+-(2SEM)0.038025488848093106      \n",
      "Epoch 186 finished      \n",
      "avg. loss: 0.13560356330126525+-(2SEM)0.03045251189282033      \n",
      "Epoch 187 finished      \n",
      "avg. loss: 0.10921160373836755+-(2SEM)0.02267385286023587      \n",
      "Epoch 188 finished      \n",
      "avg. loss: 0.10868294378742575+-(2SEM)0.035257344311621114      \n",
      "Epoch   188: reducing learning rate of group 0 to 6.1035e-08.\n",
      "Epoch 189 finished      \n",
      "avg. loss: 0.10870173890143633+-(2SEM)0.026629059917840956      \n",
      "Epoch 190 finished      \n",
      "avg. loss: 0.11700317095965147+-(2SEM)0.027558730330945923      \n",
      "Epoch 191 finished      \n",
      "avg. loss: 0.11667243205010891+-(2SEM)0.028144007464166225      \n",
      "Epoch 192 finished      \n",
      "avg. loss: 0.1339734242297709+-(2SEM)0.031461735982956165      \n",
      "Epoch 193 finished      \n",
      "avg. loss: 0.10309227034449578+-(2SEM)0.026878307547217085      \n",
      "Epoch 194 finished      \n",
      "avg. loss: 0.10739253349602222+-(2SEM)0.023193463342012367      \n",
      "Epoch 195 finished      \n",
      "avg. loss: 0.11409319434314966+-(2SEM)0.024114226105216317      \n",
      "Epoch 196 finished      \n",
      "avg. loss: 0.09083716096356512+-(2SEM)0.035493677763931725      \n",
      "Epoch 197 finished      \n",
      "avg. loss: 0.118904441408813+-(2SEM)0.0366856784945835      \n",
      "Epoch 198 finished      \n",
      "avg. loss: 0.11471632467582822+-(2SEM)0.03086726339155524      \n",
      "Epoch 199 finished      \n",
      "avg. loss: 0.11788036357611417+-(2SEM)0.024991502665943306      \n",
      "Epoch   199: reducing learning rate of group 0 to 3.0518e-08.\n",
      "Finished Training      \n"
     ]
    }
   ],
   "source": [
    "# The N2V network requires only a single output unit per pixel\n",
    "net = UNet(1, depth=3)\n",
    "\n",
    "# Split training and validation data.\n",
    "my_train_data=data[:-5].copy()\n",
    "my_val_data=data[-5:].copy()\n",
    "\n",
    "# Start training.\n",
    "trainHist, valHist = training.trainNetwork(net=net, trainData=my_train_data, valData=my_val_data,\n",
    "                                           postfix='conv_N2V', directory=path, noiseModel=None,\n",
    "                                           device=device, numOfEpochs= 200, stepsPerEpoch=10, \n",
    "                                           virtualBatchSize=20, batchSize=1, learningRate=1e-3)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "image/png": "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",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "# Let's look at the training and validation loss\n",
    "plt.xlabel('epoch')\n",
    "plt.ylabel('loss')\n",
    "plt.plot(valHist, label='validation loss')\n",
    "plt.plot(trainHist, label='training loss')\n",
    "plt.legend()\n",
    "plt.show()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.1"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 2
}