dest_onlyveg = os.path.join(outpaths['only_veg'],img_name+'.png')
dest_ndvi = os.path.join(outpaths['ndvi'],img_name+'.png')
dest_overlay = os.path.join(outpaths['overlay'],img_name+'.png')
try:
if not os.path.isfile(dest_pred):
vd.single_img_forward_pass(epoch_path,img_path)
# vd.single_img_forward_pass(epoch_path,img_path,dataset='cityscapes',model='DeepLabV3_plus')
copyfile(out_orig,dest_orig)
copyfile(out_pred,dest_pred)
if gen_extras:
cs.gen_only_veg_img(dest_pred,dest_orig,dest_onlyveg)
cs.gen_uint8_NDVI(dest_onlyveg,dest_ndvi)
cs.gen_overlay_img(dest_orig,dest_pred,dest_overlay)
except:
msc.telegram_bot_sendtext("failed on "+img_name)
# msc.telegram_bot_sendtext(e)
msc.print_rem_time_info(len(img_list),i,t1)
msc.telegram_bot_sendtext("img_gen done!!")
# cityscapes dict: /home/kaue/Semantic-Segmentation-Suite/cityscapes/class_dict.csv
classifiedList = misc.orderedFileList(classifiedPath,'*.png')
for path in classifiedList:
# print(path)
filename = misc.filenameFromPath(path)
originalImgPath = os.path.join(originalsPath,misc.modFilenameExt(filename,'.jpg'))
outPath = os.path.join(toClassifyFolder,filename)
print(originalImgPath)
classifiedImg = cv2.imread(path)
originalImg = cv2.imread(originalImgPath)
for i,column in enumerate(classifiedImg):
for j,pixel in enumerate(column):
# as there are only black or white pixels, we can test only a channel
# print(originalImg[i,j,0])
if pixel[0] == 255:
try:
pixel.itemset(0,originalImg[i,j,0])
pixel.itemset(1,originalImg[i,j,1])
pixel.itemset(2,originalImg[i,j,2])
except:
print(classifiedImg.shape)
cv2.imwrite(outPath,classifiedImg)
misc.telegram_bot_sendtext("processing ended")
pathlist = msc.getSubdirs(basedir)
# sizes = np.empty(len(pathlist), dtype=int)
# print(sizes)
# for i,path in enumerate(pathlist):
# sizes[i] = path.count('/')
# print(sizes)
# max_lvl = np.max(sizes)
# print(max_lvl)
for path in pathlist:
if msc.checkForImages(path,'*.png'):
parent = msc.get_parent_dir(path)S
imgList = msc.orderedFileList(path,'*.png')
for imgPath in imgList:
if strToContain in imgPath:
filename = msc.filenameFromPath(imgPath)
newfilename = filename.replace(strToReplace,strToAdd)
print(newfilename)
move(imgPath,os.path.join(parent,newfilename))
# os.removedirs(path)
msc.telegram_bot_sendtext("copying terminated")
from processing_functions import color_stuff as cs
from processing_functions import general_funcs as gf
import glob
import cv2
basedir = "/home/kauevestena/Dropbox/data/backup/temp/"
baseOutDir = "/home/kauevestena/Dropbox/data/gt/originals_complete_red/"
dirList = msc.getSubdirs(basedir)
subpaths = msc.subpathList(dirList,basedir)
msc.createDirStructure(baseOutDir,subpaths)
for folder in dirList:
if msc.checkForImages(folder,'*.png'):
print(folder)
subpath = msc.removeBasedir(folder,basedir)
outFolderPath = os.path.join(baseOutDir,subpath)
for filepath in glob.glob(folder+"/*.png"):
fileName = msc.filenameFromPath(filepath)
outPath = os.path.join(outFolderPath,fileName)
cs.save_one_band(filepath,outPath)
msc.telegram_bot_sendtext("band extraction terminated")
# # # # # plotting stuff
# # # # plt.close('all')
# # # # plt.figure()
# # # # picklepath = os.path.join(vd.PICKLES_PATH,key+"_"+db_name+'.pickle')
# # # # errorMetricsDfs[key].to_pickle(picklepath,compression=None)
# # # # errorMetricsDfs[key].plot.line()
# # # # plt.savefig(os.path.join(outpath,key+'.png'))
# # # # plt.close('all')
# # # # plt.figure()
# # # # errorMetricsDfs[key].T.plot.line()
# # # # plt.savefig(os.path.join(outpath,'T_'+key+'.png'))
print("tooked {} s".format(time.time() - beg))
for key in revCkptDict:
print("{} is {}".format(key, revCkptDict[key]))
revCkptDict_path = os.path.join(vd.PICKLES_PATH, 'ckpt_rev_dict.pickle')
with open(revCkptDict_path, 'wb') as pickling_dict:
pickle.dump(revCkptDict, pickling_dict)
msc.telegram_bot_sendtext("pickled and ended!!!")
#script to classify vegetation and no vegatation in a manualy classified in white for vegetation and anything to non-vegetation
from processing_functions import misc as msc
from processing_functions import color_stuff as cs
import datetime
import os
# import cv2
# inputpath = "/home/kaue/data/extracted_images/samples/2019-07-11-16-21-46/regular"
inputpath = "/home/kaue/data/extracted_images/samples/2019-07-11-16-21-46/reclassification"
# outpath = "/home/kaue/data/extracted_images/samples/2019-07-11-16-21-46/already_classified"
outpath = "/home/kaue/data/extracted_images/samples/2019-07-11-16-21-46/reclassification_blackwhite"
imagelist = msc.orderedFileList(inputpath, extension='*.png')
minDate = datetime.datetime(year=2019, day=17, month=7)
for imgpath in imagelist:
filedate = msc.getModTime(imgpath)
if filedate > minDate:
savepath = os.path.join(outpath, msc.filenameFromPath(imgpath))
# cs.remove_other_classes(imgpath,savepath,255,255,255)
cs.remove_other_classes(imgpath, savepath, 0, 0, 0, 255, 255, 255)
print(imgpath)
msc.telegram_bot_sendtext("GT classification ended")
ckptList.append(vd.checkpoint(validckptpath, "python3.7"))
else:
print(validckptpath)
ckptList.append(vd.checkpoint(validckptpath))
# for ckpt in ckptList:
# print(ckpt.model,ckpt.dataset)
# print(listOfListsOnDirs[0][0])
# ckptList[0].process_image(gtImagesList[0])
# ckptList[0].process_and_validate(imgs_and_gts[0])
# ckptList[0].process_and_validate(imgs_and_gts[1])
# print(ckptList[0].error_metrics_store)
# ckptList[0].dump_to_tinyDB()
print(len(ckptList), len(imgs_and_gts))
total_its = len(ckptList) * len(imgs_and_gts)
for i, ckpt in enumerate(ckptList):
for j, img in enumerate(imgs_and_gts):
beg = time.time()
ckpt.process_and_validate(img, True, False)
msc.print_rem_time_info(total_its, i * len(imgs_and_gts) + j, beg)
ckpt.dump_to_tinyDB()
msc.telegram_bot_sendtext("evaluation done")