From 5ccd9f5b66b2f2d0887bfa3d6bd27af489c001db Mon Sep 17 00:00:00 2001 From: fdreyer Date: Thu, 8 Aug 2019 16:56:32 +0100 Subject: [PATCH] changed model plotting routine --- scripts/plotting/plot_model.py | 96 +++++++++++++++++++++++++++++----- 1 file changed, 83 insertions(+), 13 deletions(-) diff --git a/scripts/plotting/plot_model.py b/scripts/plotting/plot_model.py index 5e797cb..ec6acc4 100755 --- a/scripts/plotting/plot_model.py +++ b/scripts/plotting/plot_model.py @@ -3,36 +3,107 @@ import argparse, yaml, pprint, os, shutil, datetime, sys, pickle import numpy as np +import matplotlib.pyplot as plt +from matplotlib.backends.backend_pdf import PdfPages +from matplotlib.patches import FancyArrowPatch from cyclejet.cyclegan import CycleGAN -from cyclejet.tools import loss_calc, plot_model +from cyclejet.tools import loss_calc, plot_model, xval, yval from cyclejet.scripts.run import load_yaml +from random import randrange + +def plot_event(fn, refA, refB, predictA, predictB, + predictA2, predictB2, averager, titleA=None, titleB=None): + with PdfPages(fn) as pdf: + fig, axs = plt.subplots(2,3, figsize=(7.5,5.5)) + plt.subplots_adjust(wspace=0.5,hspace=0.52) + i = randrange(len(refA)) + # figtr = fig.transFigure.inverted() + # ptB = figtr.transform(ax0tr.transform((225., -10.))) + # ptE = figtr.transform(ax1tr.transform((225., 1.))) + # arrow=FancyArrowPatch( + # ptB, ptE, transform=fig.transFigure, # Place arrow in figure coord system + # fc = "g", connectionstyle="arc3,rad=0.2", arrowstyle='simple', alpha = 0.3, + # mutation_scale = 40.) + # fig.patches.append(arrow) + axs[0,0].imshow(refA[i].transpose(),vmin=0.0,vmax=0.2,origin='lower', + aspect='auto', extent=[xval[0], xval[1], yval[0], yval[1]]) + axs[0,0].set_title('A' if not titleA else titleA) + axs[0,0].set_xticks([]) + axs[0,0].set_yticks([]) + axs[0,1].imshow(predictB[i].transpose(),vmin=0.0,vmax=0.2,origin='lower', + aspect='auto', extent=[xval[0], xval[1], yval[0], yval[1]]) + # axs[0,1].set_xlabel('$\ln(1 / \Delta_{ab})$') + # axs[0,1].set_ylabel('$\ln(k_{t} / \mathrm{GeV})$',labelpad=-2) + axs[0,1].set_title('B' if not titleB else titleB) + axs[0,1].set_xticks([]) + axs[0,1].set_yticks([]) + axs[0,2].imshow(predictB2[i].transpose(),vmin=0.0,vmax=0.2,origin='lower', + aspect='auto', extent=[xval[0], xval[1], yval[0], yval[1]]) + axs[0,2].set_title('A' if not titleA else titleA) + + axs[0,2].set_xticks([]) + axs[0,2].set_yticks([]) + axs[1,0].imshow(averager.inverse(refA)[i].transpose(), + vmin=0.0,vmax=0.2,origin='lower',aspect='auto', + extent=[xval[0], xval[1], yval[0], yval[1]]) + #axs[1,0].set_title('A' if not titleA else titleA) + axs[1,0].set_xticks([]) + axs[1,0].set_yticks([]) + axs[1,1].imshow(averager.inverse(predictB)[i].transpose(), + vmin=0.0,vmax=0.2,origin='lower',aspect='auto', + extent=[xval[0], xval[1], yval[0], yval[1]]) + # axs[0,1].set_xlabel('$\ln(1 / \Delta_{ab})$') + # axs[0,1].set_ylabel('$\ln(k_{t} / \mathrm{GeV})$',labelpad=-2) + #axs[1,1].set_title('B' if not titleB else titleB) + axs[1,1].set_xticks([]) + axs[1,1].set_yticks([]) + axs[1,2].imshow(averager.inverse(predictB2)[i].transpose(), + vmin=0.0,vmax=0.2,origin='lower',aspect='auto', + extent=[xval[0], xval[1], yval[0], yval[1]]) + #axs[1,2].set_title('A' if not titleA else titleA) + axs[1,2].set_xticks([]) + axs[1,2].set_yticks([]) + # pdf.savefig() + # plt.close() + + # fig, axs = plt.subplots(3, 2, figsize=(6,8)) + plt.close() + pdf.savefig(fig) + def main(args): - if os.path.isfile(args.model.strip('/')+'/best-model.yaml'): - fn=args.model.strip('/')+'/best-model.yaml' + model=args.model.strip('/') + if os.path.isfile(model+'/best-model.yaml'): + fn=model+'/best-model.yaml' else: - fn=args.model.strip('/')+'/input-runcard.json' + fn=model+'/input-runcard.json' hps=load_yaml(fn) cgan = CycleGAN(hps) - cgan.load(args.model.strip('/')) + cgan.load(model) refA=np.array(cgan.imagesA) refB=np.array(cgan.imagesB) # generating predicted sample predictA=cgan.g_BA.predict(refA) predictB=cgan.g_AB.predict(refB) + predictA2=cgan.g_AB.predict(predictA) + predictB2=cgan.g_BA.predict(predictB) refA = cgan.preproc.inverse(refA) refB = cgan.preproc.inverse(refB) predictA = cgan.preproc.inverse(predictA) predictB = cgan.preproc.inverse(predictB) if args.savefull: - np.save('%s/referenceA'%args.model.strip('/'), refA) - np.save('%s/referenceB'%args.model.strip('/'), refB) - np.save('%s/predictedA'%args.model.strip('/'), predictA) - np.save('%s/predictedB'%args.model.strip('/'), predictB) + np.save('%s/referenceA'%model, refA) + np.save('%s/referenceB'%model, refB) + np.save('%s/predictedA'%model, predictA) + np.save('%s/predictedB'%model, predictB) - # now create diagnostic plots - figfn='%s/result.pdf' % args.model.strip('/') - plot_model(figfn, refA, refB, predictA, predictB, + # now create plots + figfn1='%s/result.pdf' % model + plot_model(figfn1, refA, refB, predictA, predictB, + titleA=args.titleA, titleB=args.titleB) + figfn2='%s/result_event.pdf' % model + plot_event(figfn2, refA, refB, predictA, predictB, + predictA, predictB, cgan.avg, titleA=args.titleA, titleB=args.titleB) #---------------------------------------------------------------------- @@ -46,7 +117,6 @@ def main(args): help='Title of sample A.') parser.add_argument('--titleB', type=str, default=None, help='Title of sample A.') - parser.add_argument('--savefull', action='store_true') args = parser.parse_args() main(args)