def fetch_image_urls(path):
data_dictionary = []
data_urls = []
tl_list = []
br_list = []
raw_data = read_json_file(path)
for idx, data_dict in enumerate(raw_data):
data_dict, img_tl_list, img_br_list = generate_tl_br_list(
data_dict=data_dict, data_urls=data_urls)
data_dictionary.append(data_dict)
tl_list.append(img_tl_list)
br_list.append(img_br_list)
objects = []
for i in range(len(data_dict.get('annotation'))):
objects.append(data_dict.get('annotation')[i].get('label')[0])
write_xml(folder='./dataset',
img_path='./dataset/image{}.jpg'.format(str(idx)),
img_name='image{}.jpg'.format(str(idx)),
objects=objects,
tl=tl_list[idx],
br=br_list[idx],
savedir='./annotation')
print("[INFO]: Xml files has been generated!")
if os.path.getsize('./image_urls.txt') == 0:
write_image_urls_to_file(data_urls)
return data_dictionary, data_urls, tl_list, br_list
def onkeypress(event):
global object_list
global tl_list
global br_list
global img
if event.key == 'r':
tl_list = []
br_list = []
object_list = []
print("Object List cleared")
# if event.key == 'r':
# print(object_list)
# write_xml(image_folder, img, object_list, tl_list, br_list, savedir)
# tl_list = []
# br_list = []
# object_list = []
# img = None
# keyboard = Controller()
# keyboard.press('q')
# keyboard.release('q')
if event.key == 't':
print(object_list)
write_xml(image_folder, img, object_list, tl_list, br_list, savedir,
sys.argv[1])
img = None
keyboard = Controller()
keyboard.press('q')
keyboard.release('q')
def onekeypress(event):
global br_list
global tl_list
global object_list
global img
if event.key == 'q':
write_xml(image_folder, img, object_list, tl_list, br_list, savedir)
br_list = []
tl_list = []
object_list = []
img = None
plt.close()
def onkeypress(event):
global object_list
global tl_list
global br_list
global img
if event.key == 'q':
print(object_list)
write_xml(image_folder, img, object_list, tl_list, br_list, savedir)
tl_list = []
br_list = []
object_list = []
img = None
def onkeypress(event):
global object_list
global tl_list
global br_list
global img
if event.key == 'q':
write_xml(image_folder, img, object_list, tl_list, br_list, savedir)
# 변수 초기화 & 이미지 닫기
tl_list = []
br_list = []
object_list = []
img = None
def onkeypress(event):
global object_list
global tl_list
global br_list
global img
if event.key == "q":
print(object_list)
write_xml(IMAGE_FOLDER, img, object_list, tl_list, br_list, SAVE_DIR)
tl_list = []
br_list = []
object_list = []
img = None
plt.close()
def onkeypress(event):
global object_list
global top_left_list
global bottom_right_list
global img
if event.key == 'q':
print(top_left_list, bottom_right_list)
write_xml(image_folder, path, object_list, top_left_list,
bottom_right_list, savedir)
top_left_list = []
bottom_right_list = []
object_list = []
plt.close()
def makeAnnos(lego, top_l_arr, name, width, height, n):
out_top_l = []
out_bot_r = []
for l in range(len(lego)):
xmin, ymin = getEdges(lego[l])
top_l = top_l_arr[l]
stop_l = top_l
top_l = (top_l[0] + xmin[0] - 5, top_l[1] + ymin[0] - 5)
bot_r = (stop_l[0] + xmin[1] + 5, stop_l[1] + ymin[1] + 5)
out_top_l.append(top_l)
out_bot_r.append(bot_r)
write_xml("Images", [width, height, 3], n, out_top_l, out_bot_r,
savedir_anno, name)
def on_key_press(event):
global object_list
global top_left_list
global bottom_right_list
global img
if event.key == 'q':
write_xml(image_folder, img, object_list, top_left_list,
bottom_right_list, save_directory)
# print(top_left_list, bottom_right_list)
top_left_list = []
bottom_right_list = []
object_list = []
img = None
plt.close()
def onkeypress(event):
global object_list
global tl_list
global br_list
global img
if event.key == "q":
write_xml(data_folder, img, object_list, tl_list, br_list, savedir,
"jpg")
object_list = []
tl_list = []
br_list = []
img = None
plt.close()
def onkeypress(event):
# TODO Implement multi object declaration (Change savedir)
global object_list
global tl_list
global br_list
global img
if event.key == 'q':
print(object_list)
write_xml('Tes_Model/images', img, object_list, tl_list, br_list, savedir)
tl_list = []
br_list = []
object_list = []
img = None
plt.close()
def onkeypress(event):
global object_list
global tl_list
global br_list
global img
global save
global objType
global isSaved
if event.key == 'c':
objType = objType + 1
if objType == 2:
objType = 0
print("Current type:" + classes[int(objType)])
if event.key == 'b':
save = True
if event.key == 't':
tl_list = []
br_list = []
object_list = []
print("Resetted")
if event.key == 'q':
print(tl_list, br_list, object_list)
#write_txt_file(tl_list, br_list, object_list)
if len(object_list) > 0:
write_xml(image_folder, img, object_list, tl_list, br_list,
savedir, 0)
mirror(image_folder, savedir, img, tl_list, br_list, object_list)
scale(image_folder, savedir, img, tl_list, br_list, object_list)
translation(image_folder, savedir, img, tl_list, br_list,
object_list)
print("Skrev til xml")
tl_list = []
br_list = []
object_list = []
img = None
plt.close()
if event.key == 'm':
plt.close()
#if not isSaved:
#write_txt_file(tl_list, br_list, object_list)
global running
global annotation_file
#annotation_file.close()
running = False
def onkeypress(event):
global object_list
global tl_list
global br_list
global img
if event.key == 'a': #press p if it's a person
object_list.append(obj1)
write_xml(image_folder, img, object_list,tl_list,br_list,savedir)
print(tl_list,br_list,object_list)
#write_xml(image_folder, img, object_list, tl_list, br_list, savedir)
tl_list = []
br_list = []
object_list = []
img = None
plt.close()
'''
def onkeypress(event):
global object_list
global tl_list
global br_list
global img,obj
if event.key == 'q':
print(object_list)
write_xml(image_folder, img, object_list, tl_list, br_list, savedir)
tl_list = []
br_list = []
object_list = []
img = None
plt.close()
if event.key in ["1","2","3","4","5","6","7","8","9"]:
obj = objlist[int(event.key )-1]
print(event.key,obj)
def translation(image_src, xml_path, image, tl, br, obj):
image_cv = cv2.imread(image.path)
tl_translated = []
br_translated = []
height, width, depth = image_cv.shape
#translation factors
factor = 0.2
tx = random.randint(-int(width * factor), int(width * factor))
ty = random.randint(-int(height * factor), int(height * factor))
#Translation matrix
M = np.float32([[1, 0, tx], [0, 1, ty]])
translated_image = cv2.warpAffine(image_cv, M, (width, height))
print("%f %f" % (tx, ty))
for i in range(0, len(tl)):
if ((tx + tl[i][0]) > width and
(tx + br[i][0]) > width) or ((tx + tl[i][0]) < 0 and
(tx + br[i][0]) < 0):
print("bb er utenfor x-omraade")
continue
if ((ty + tl[i][1]) > height and
(ty + br[i][1]) > height) or ((ty + tl[i][1]) < 0 and
(ty + br[i][1]) < 0):
print("bb er utenfor y-omraade")
continue
tl_translated.append(
((tx + tl[i][0] if
(tx + tl[i][0]) > 0 else 0), (ty + tl[i][1] if
(ty + tl[i][1] > 0) else 0)))
br_translated.append(
((tx + br[i][0] if (tx + br[i][0]) < width else width),
(ty + br[i][1] if (ty + br[i][1] < height) else height)))
print("tl_translated", tl_translated)
print("br_translated", br_translated)
#cv2.rectangle(translated_image, tl_translated[i], br_translated[i], (255, 0, 0), 1)
write_xml(image_src, image, obj, tl_translated, br_translated, xml_path, 3)
cv2.imwrite(image_src + "/" + "t" + image.name, translated_image)
def onkeypress(event):
global object_list
global tl_list
global br_list
global img
if event.key == 'q':
write_xml(target_folder_gray, img, object_list, tl_list, br_list,
savedir)
print(f'[Final] \n{tl_list}\n{br_list}\n{object_list}\n')
tl_list = []
br_list = []
img = None
object_list = []
plt.close()
if event.key == 'a':
os.remove(img)
print('[Deleted the image]\n')
plt.close()
def mirror(image_src, xml_path, image, tl, br, obj):
image_cv = cv2.imread(image.path)
image_flipped = cv2.flip(image_cv, 1)
tl_mirrored = []
br_mirrored = []
height, width, depth = image_cv.shape
print(len(tl))
for i in range(0, len(tl)):
tl_mirrored.append((width - br[i][0], tl[i][1]))
br_mirrored.append((width - tl[i][0], br[i][1]))
#cv2.rectangle(image_flipped, tl_mirrored[i], br_mirrored[i], (255, 0, 0), 1)
cv2.imwrite(image_src + "/" + "m" + image.name, image_flipped)
#Safeguard to prevent xml file writing with a non existent image
if os.path.isfile(image_src + "/" + "m" + image.name):
print("writing translated xml_file")
write_xml(image_src, image, obj, tl_mirrored, br_mirrored, xml_path, 1)
def onkeypress(event):
global object_list
global tl_list
global br_list
global img
if event.key == 'q': #Next Image
print(object_list)
if(tl_list and br_list)!=[]:
write_xml(image_folder, img, object_list, tl_list, br_list, savedir)
tl_list = []
br_list = []
object_list = []
img = None
else:
object_list = []
img = None
if event.key=='w': #to clear any box drawn by mistake.
tl_list = []
br_list = []
object_list = []
def on_key_press(event):
global object_list
global tl_list
global br_list
global img
if event.key == "q":
write_xml("imgData", img, object_list, tl_list, br_list,
path + save_dir)
os.rename(img.path, "../resources/imgData/" + img.name)
tl_list = []
br_list = []
object_list = []
img = None
plt.close()
if event.key == "w":
os.remove(img.path)
tl_list = []
br_list = []
object_list = []
img = None
plt.close()
def onkeypress(event):
global object_list
global tl_list
global br_list
global tl
global br
global img
if event.key == 'q':
print(object_list)
tl.append(tl_list[-1])
br.append(br_list[-1])
write_xml(image_folder, img, object_list, tl, br, savedir)
tl_list = []
br_list = []
object_list = []
tl = []
br = []
img = None
if event.key == 'w':
tl.append(tl_list[-1])
br.append(br_list[-1])
def scale(image_src, xml_path, image, tl, br, obj):
image_cv = cv2.imread(image.path)
tl_scaled = []
br_scaled = []
#Scaling factors
fx = random.uniform(0.75, 1.25)
fy = random.uniform(0.75, 1.25)
scaled_img = cv2.resize(image_cv,
None,
fx=fx,
fy=fy,
interpolation=cv2.INTER_CUBIC)
print("%f %f" % (fx, fy))
for i in range(0, len(tl)):
tl_scaled.append((int(fx * tl[i][0]), int(fy * tl[i][1])))
br_scaled.append((int(fx * br[i][0]), int(fy * br[i][1])))
#cv2.rectangle(scaled_img, tl_scaled[i], br_scaled[i], (255, 0, 0), 1)
write_xml(image_src, image, obj, tl_scaled, br_scaled, xml_path, 2)
cv2.imwrite(image_src + "/" + "s" + image.name, scaled_img)
pt1 = (xmin, ymin)
pt2 = (xmax, ymax)
cv2.rectangle(image, pt1, pt2, (255, 255, 0), 3)
loc = (xmin, ymin - 20)
cv2.putText(image, pid, loc, cv2.FONT_HERSHEY_SIMPLEX, 1.5,
(255, 0, 255), 6)
#The frame is ready and already captured
# cv2.namedWindow('video', cv2.WINDOW_NORMAL)
# cv2.imshow('video', image)
# if cv2.waitKey(1) & 0xFF == ord('q'):
# break
videowriter.write(image)
bbox.append((xmin, ymin, xmax, ymax))
objects.append(str(pid))
# f = open(write_fname,'a')
# txt = pid + ' ' + str(xmin) + ' ' + str(ymin) + ' ' + str(xmax) + \
# ' ' + str(ymax) + ' '
# f.write(txt)
# f.write('\n')
filename = 'frame_no_' + '{0:05}'.format(frame_no) + '.jpg'
write_xml(folder, orig_image, filename, path, objects, bbox, savedir)
print("--- %s seconds ---" % (time.time() - start_time))
cap.release()
videowriter.release()
#cv2.destroyAllWindows()
) #Arka plan üzerine koyulan resim için rastgele bir satır belirleniyor ancak bu sayı merkezi temsil ettiği için arkaplanın maksimum yüksekliğinden yerleştirilen resmin yüksekliğinin yarısı çıkartılıyor.
rndCol = rd.randint(0,
(Backgroundimage.shape[1] - Croppedimage.shape[1]
)) #yukarıdaki işlem sütun için tekrarlanıyor.
img1[
rndRow:rows + rndRow, rndCol:cols +
rndCol] = dst #Kesilmiş resmin arkaplan üzerine koyulması tamamlandı.
#cv2.imshow('res', img1)
cv2.imwrite(dataset_folder_path + 'FinalImages\\{:06}.png'.format(i),
img1) #resimler altı haneli bir formatta kaydediliyor.
br_list.append(
(int(rndRow + (Croppedimage.shape[0] / 2)),
(int(rndCol + (Croppedimage.shape[1] / 2))))
) #rastgele belirlenen merkez noktası üzerinden aşağı sağ(bottom right br) ve yukarı sol (top left tl) noktaları hesaplanıyor
tl_list.append(
(int(rndRow - (Croppedimage.shape[0] / 2)),
(int(rndCol - (Croppedimage.shape[1] / 2))
))) #hesaplanan bu noktalar xml oluştururken kullanılacak
object_list.append(
CropfileName.name
) #Klasör ismi aynı zaman içerisindeki resimlerin eğitim türü adı oluyor
write_xml(
image_folder,
dataset_folder_path + 'FinalImages\\{:06}.png'.format(i),
'{:06}.png'.format(i), object_list, tl_list, br_list, savedir
) #xml oluşturma komutu. Bu komut ile annotation dosyaları oluşuyor.
tl_list.clear() #listeler sıfırlanıyor
br_list.clear()
object_list.clear()
i = i + 1 #isimlendirmek için kullanılan sayaç bir artırılıyor.
