def jetson_nano_request(product_id, username, image_name, image_src, ip,
function):
"""
目前只支援jetson nano /bread/query/basic/v2 api
將圖片資訊發送request至jetson nano做圖片偵測擷取其中物件
並將結果的YOLO model output id 轉換為客戶product id
儲存至convert_recognition.data 及 recognition.data
:param product_id: 客戶產品ID
:param username: 此專案都預設為ROOT
:param image_name: 該圖片的檔案名稱
:param image_src: 圖片轉成base64字串
:param ip: Jetson nano ip
:param function: Jetson nano 提供的API URL
:return:
"""
# image_src is only one image base64
send_data = {
"function": "basic",
"values": [{
"id": 0,
"b64string": image_src
}]
}
try:
# for testing , fake API response
ans = {
"rtn":
200,
"items": [{
"id":
0,
"item": [{
"num": 0,
"position": [200, 200, 200, 200],
"confidence": 0.8,
"real_id": 1,
"name": "Taro"
}]
}]
}
if not TESTING_NOT_CONNET_JETSON_NANO:
ans = requests.post(ip + function, json=send_data).json()
except Exception as e:
# request fail
print("%s %s location recognition request fail" %
(product_id, image_name))
return ""
if ans['rtn'] == 400:
print("%s %s location recognition request fail x" %
(product_id, image_name))
return ""
# Parsing data to dict
convert_dict = {}
response_data = ans["items"]
for data in response_data:
id = str(data["item"][0]["real_id"])
if id not in convert_dict:
convert_dict[id] = {product_id: 1}
else:
if product_id in convert_dict[id]:
convert_dict[id][product_id] += 1
else:
convert_dict[id][product_id] = 1
################ update convert_recognition.data################
convert_json_path = CUSTOMER_FILE + username + '/convert_recognition.data'
data_lock = CUSTOMER_FILE + username + '/convert_datalock'
# if data lock exited , sleep a little time
while isfile(data_lock):
lockfile_sleep(data_lock)
# create data lock file
create_lock_file(data_lock)
# if convert_recognition.data not exited , create it
if not isfile(convert_json_path):
init_json_file(convert_json_path)
convert_recognition_dict = get_json_data(convert_json_path)
convert_recognition_dict[product_id + '/' + image_name] = convert_dict
overwrite_json_data(convert_json_path, convert_recognition_dict)
# delete data lock file
delete_lock_file(data_lock)
# update recognition.data
recognition_json_path = CUSTOMER_FILE + username + '/recognition.data'
data_lock = CUSTOMER_FILE + username + '/recognition_datalock'
# if data lock exited , sleep a little time
while isfile(data_lock):
lockfile_sleep(data_lock)
# create data lock file
create_lock_file(data_lock)
# if recognition.data not exited , create it
if not isfile(recognition_json_path):
init_json_file(recognition_json_path)
recognition_dict = get_json_data(recognition_json_path)
recognition_dict_keys = recognition_dict.keys()
keys = convert_dict.keys()
for key in keys:
if key in recognition_dict:
if product_id in recognition_dict[key]:
recognition_dict[key][product_id] += convert_dict[key][
product_id]
else:
recognition_dict[key][product_id] = convert_dict[key][
product_id]
else:
recognition_dict[key] = convert_dict[key]
overwrite_json_data(recognition_json_path, recognition_dict)
# delete data lock file
delete_lock_file(data_lock)
def delete_product_image():
"""
儲存產品圖片，可以一次刪除多張圖片
刪除圖片後開啟recognition.data, convert_recognition.data 清除該圖片對應映射資料
"""
if request.method == "POST":
username = '******'
data = request.get_json()
image_name = data["file_path"]
product_id = str(data["product_id"])
path = CUSTOMER_FILE + username + '/'
image_path = path + product_id + '/' + image_name
data_lock = path + '/datalock'
####### delete image file and rewrite json data image count
while isfile(data_lock):
lockfile_sleep(data_lock)
# create data lock file
with open(data_lock, 'w') as lock:
pass
try:
remove(image_path)
except Exception as e:
return {"message": "image delete error"}
json_file_path = path + product_id + '/' + 'data.data'
rewrite_count = str(int(get_json_image_count(json_file_path)) - 1)
rewrite_json_image_count(json_file_path, rewrite_count)
delete_lock_file(data_lock)
##############################################################
####### rewrite recognition.data and convert_recognition.data
convert_json_path = CUSTOMER_FILE + username + '/convert_recognition.data'
data_lock = CUSTOMER_FILE + username + '/convert_datalock'
convert_json_key = product_id + '/' + image_name
while isfile(data_lock):
lockfile_sleep(data_lock)
# create data lock file
create_lock_file(data_lock)
convert_recognition_dict = get_json_data(convert_json_path)
# get data which key is convert_json_key and delete the key
data_dict = convert_recognition_dict.pop(convert_json_key)
overwrite_json_data(convert_json_path, convert_recognition_dict)
# delete data lock file
delete_lock_file(data_lock)
#
recognition_json_path = CUSTOMER_FILE + username + '/recognition.data'
data_lock = CUSTOMER_FILE + username + '/recognition_datalock'
while isfile(data_lock):
lockfile_sleep(data_lock)
# create data lock file
create_lock_file(data_lock)
recognition_dict = get_json_data(recognition_json_path)
model_keys = data_dict.keys()
for model_key in model_keys:
product_keys = data_dict[model_key].keys()
for product_key in product_keys:
recognition_dict[model_key][product_key] -= data_dict[
model_key][product_key]
overwrite_json_data(recognition_json_path, recognition_dict)
# delete data lock file
delete_lock_file(data_lock)
print("delete image all process done!")
return {"message": "done"}
return {"message": "post only"}
def save_product_image():
"""
儲存產品圖片，可以是多張圖片一次傳入
所有的圖片只會有第一張圖片被送到jetson nano 辨識物件
這是為了避免jetson nano 效能被操到當機
"""
if request.method == "POST":
username = USERFILE
data = request.get_json()
img_srcs = data["base64"]
product_id = str(data["product_id"])
first_image_name = []
flag = 0
for i, img_src in enumerate(img_srcs):
data_lock = CUSTOMER_FILE + username + '/' + product_id + '/datalock'
# if data lock exited , sleep a little time
while isfile(data_lock):
lockfile_sleep(data_lock)
# create data lock file
with open(data_lock, 'w') as lock:
pass
# save image
path = CUSTOMER_FILE + username + '/' + product_id
files = listdir(path)
# final_number = '1'
image_count = 0
# check out file image number to decide savefile name
for i, file in enumerate(files):
data_type = file.split(".")[-1]
if data_type == "jpg" and file.split(".")[0].split(
"_")[-1] != 'title':
image_count += 1
# final_number = str(int(file.split(".")[0].split("_")[-1])+1)
save_data_name = generate_random_str(25) + '.jpg'
if flag == 0:
flag += 1
first_image_name.append(save_data_name)
try:
img = imgSrc2_cv2img(img_src)
cv2.imwrite(path + '/' + save_data_name, img,
[cv2.IMWRITE_JPEG_QUALITY, 90])
except Exception as e:
delete_lock_file(data_lock)
return {"message": "image save error"}
# overwrite data.data
json_file_path = path + '/' + 'data.data'
image_count += 1
rewrite_json_image_count(json_file_path, str(image_count))
delete_lock_file(data_lock)
# send request to jetson nano
t = threading.Thread(target=jetson_nano_request,
args=(product_id, username, first_image_name[0],
img_srcs[0], TEST_JETSONNANO_IP,
JETSONNANO_API_BASIC))
t.start()
return {"message": "done"}
return {"message": "post only"}
def customer_delete_product_data():
"""
客戶商品清單頁面選取商品刪除觸發API , 可以一次刪除多筆資料
:return: {"message": str, "status": str}
message:
"""
if request.method == "DEL":
username = '******'
data = request.get_json()
# get product id as list
product_ids = data["product_id"]
# 尋找刪除的productID在recognition.data 與 convert_recognition.data中的資料並刪除
# get json data from convert_recognition.data
recognition_json_path = CUSTOMER_FILE + username + '/recognition.data'
convert_json_path = CUSTOMER_FILE + username + '/convert_recognition.data'
data_lock = CUSTOMER_FILE + username + '/convert_datalock'
# if not isfile(convert_json_path) or not isfile(recognition_json_path):
# return {"message": "not image data", "status": "02"}
# if data lock exited , sleep a little time
while isfile(data_lock):
lockfile_sleep(data_lock)
# create data lock file
create_lock_file(data_lock)
convert_recognition_dict = get_json_data(convert_json_path)
convert_recognition_dict_keys = convert_recognition_dict.keys()
delete_covert_keys = []
for key in convert_recognition_dict_keys:
pro_key = key.split('/')[0]
if pro_key in product_ids:
delete_covert_keys.append(key)
for key in delete_covert_keys:
del convert_recognition_dict[key]
# convert_recognition data done
# make recognition
overwrite_recognition_dict = {}
convert_keys = convert_recognition_dict.keys()
for convert_key in convert_keys:
recognition_keys = convert_recognition_dict[convert_key].keys()
for recognition_key in recognition_keys:
if recognition_key not in overwrite_recognition_dict:
overwrite_recognition_dict[recognition_key] = \
convert_recognition_dict[convert_key][recognition_key]
else:
product_keys = convert_recognition_dict[convert_key][
recognition_key].keys()
for product_key in product_keys:
if product_key not in overwrite_recognition_dict[
recognition_key]:
overwrite_recognition_dict[recognition_key][product_key] = \
convert_recognition_dict[convert_key][recognition_key][product_key]
else:
overwrite_recognition_dict[recognition_key][product_key] += \
convert_recognition_dict[convert_key][recognition_key][product_key]
overwrite_json_data(convert_json_path, convert_recognition_dict)
# delete data lock file
delete_lock_file(data_lock)
data_lock = CUSTOMER_FILE + username + '/recognition_datalock'
while isfile(data_lock):
lockfile_sleep(data_lock)
# create data lock file
create_lock_file(data_lock)
# overwirte recognition.data
overwrite_json_data(recognition_json_path, overwrite_recognition_dict)
# delete data lock file
delete_lock_file(data_lock)
# delete file
for product_id in product_ids:
file_path = CUSTOMER_FILE + username + '/' + product_id
try:
rmtree(file_path)
except Exception as e:
print(e)
return {"message": "done", "status": "01"}
return {"message": "del only", "status": "02"}
def Recognition_bread():
"""
發送Post Request至Jetson nano上,API格式查詢gitlab 機器設備issues
目前頁面拔除該功能, 此API暫時用不到
data : JSON "b64string": 圖片的base64
:return:
"""
if request.method == "POST":
username = USERFILE
data = request.get_json()
# for testing image shape
send_data = {
"function": "basic",
"values": [{
"id": 0,
"b64string": data["b64string"]
}]
}
try:
# for testing , fake API response
response_data = {
"rtn":
200,
"items": [{
"id":
0,
"item": [{
"num": 0,
"position": [200, 200, 200, 200],
"confidence": 0.8,
"real_id": 1,
"name": "Taro"
}]
}]
}
if not TESTING_NOT_CONNET_JETSON_NANO:
ans = requests.post(TEST_JETSONNANO_IP + JETSONNANO_API_BASIC,
json=send_data)
response_data = ans.json()
print("response_data", response_data)
recognition_json_path = CUSTOMER_FILE + username + '/recognition.data'
data_lock = CUSTOMER_FILE + username + '/recognition_datalock'
# if data lock exited , sleep a little time
while isfile(data_lock):
lockfile_sleep(data_lock)
# create data lock file
create_lock_file(data_lock)
recognition_dict = get_json_data(recognition_json_path)
# delete data lock file
delete_lock_file(data_lock)
prepare_data = []
for index, data_item in enumerate(
response_data["items"][0]["item"]):
# model_id is response yolo model output
model_id = str(data_item["real_id"]) #########
print("model_id", model_id)
max_id = ''
if model_id in recognition_dict:
max_count = 0
for key in recognition_dict[model_id].keys():
if recognition_dict[model_id][key] > max_count:
max_count = recognition_dict[model_id][key]
max_id = key
if max_id == '':
max_id = random_product_id(username)
data_lock = CUSTOMER_FILE + username + '/' + max_id + '/datalock'
# create data lock file
create_lock_file(data_lock)
data_json = get_json_data(CUSTOMER_FILE + username + '/' +
max_id + '/data.data')
# delete data lock file
delete_lock_file(data_lock)
data_name = data_json["product_name"]
prepare_data.append([max_id, data_name])
for i, datas in enumerate(prepare_data):
response_data["items"][0]["item"][i]["real_id"] = datas[0]
response_data["items"][0]["item"][i]["name"] = datas[1]
return response_data
except Exception as e:
print(e)
return {"rtn": 400}
else:
return {"rtn": 400, "message": "Error, pls using post"}
def roiInference():
"""
從Camera辨識發起辨識 request 將前端camera資訊帶入後端,
前端網頁img介面為400*400但傳送的base64為原圖尺寸, 需要將ROI的4個定位點依照尺寸的差異比例進行縮放,
處理完畢後將資料打入jetson nano API /bread/query/nine/v2中,
收到response後將內容整理成html以及資料特定格式回傳至前端做顯示
:return:
response_data = {
"html": list,
"items": list,
"position": list
}
html : 交給JS渲染網頁
position: 經由後端處理將重複的ID整合交給JS渲染個別商品的ROI
items: jetson nano API /bread/query/nine/v2 的response
"""
if request.method == "POST":
username = USERFILE
request_data = request.get_json()
datas = request_data["data"]
print('get input datas:', datas)
# keep_data = [[key,AnchorWidth, AnchorHeight],[key,AnchorWidth, AnchorHeight],....]
keep_data = []
datas_jetson = []
# 前端網頁img介面為400*400但傳送的base64為原圖尺寸, 需要將ROI的4個定位點依照尺寸的差異比例進行縮放,
for i, re_data in enumerate(datas):
keep_data.append([
re_data["key"], re_data["AnchorWidth"], re_data["AnchorHeight"]
])
ip = re_data["ip"]
painted = re_data["painted"]
coordinates = []
if painted:
x = int(re_data["coordinate"][0][0] * re_data["AnchorWidth"])
y = int(re_data["coordinate"][0][1] * re_data["AnchorHeight"])
coordinates.append([x, y])
x = int(re_data["coordinate"][1][0] * re_data["AnchorWidth"])
y = int(re_data["coordinate"][1][1] * re_data["AnchorHeight"])
coordinates.append([x, y])
x = int(re_data["coordinate"][2][0] * re_data["AnchorWidth"])
y = int(re_data["coordinate"][2][1] * re_data["AnchorHeight"])
coordinates.append([x, y])
x = int(re_data["coordinate"][3][0] * re_data["AnchorWidth"])
y = int(re_data["coordinate"][3][1] * re_data["AnchorHeight"])
coordinates.append([x, y])
datas_jetson.append({
"id": i,
"cam_type": "ip_cam",
"cam_addr": ip,
"roi": coordinates
})
send_data = {"function": "nine", "values": datas_jetson}
# try:
# 發送給request 給 jestson nano /bread/query/nine/v2
if not TESTING_NOT_CONNET_JETSON_NANO:
ans = requests.post(TEST_JETSONNANO_IP + JETSONNANO_API_NINE,
json=send_data)
response_data = ans.json()
print()
else:
# 僅供測試用， TESTING_NOT_CONNET_JETSON_NANO 為TRUE的話不發送request 到jetson nano
base64_test = ""
response_data = {
"rtn":
200,
"items": [{
"id":
0,
"b64string":
base64_test,
"roi":
datas_jetson[0]["roi"],
"percentage":
0.5,
"item": [{
"num":
0,
"position": [[-8, -4], [261, -4], [261, 454],
[-8, 454]],
"confidence":
0.8,
"real_id":
2,
"name":
"BA"
}]
}, {
"id":
1,
"b64string":
base64_test,
"roi":
datas_jetson[1]["roi"],
"percentage":
0.7,
"item": [{
"num":
0,
"position": [[30, 170], [1025, 400], [50, 150],
[30, 70]],
"confidence":
0.8,
"real_id":
1,
"name":
"Taro"
}]
}]
}
# response rtn欄位不等於200表示錯誤狀態 直接回傳前端錯誤結果
if response_data["rtn"] != 200:
return response_data
# 開始將原yolo model output id 轉換為customer product id
# 從 recognition.data 取得 customer product 映射
recognition_json_path = CUSTOMER_FILE + username + '/recognition.data'
data_lock = CUSTOMER_FILE + username + '/recognition_datalock'
# if data lock exited , sleep a little time
while isfile(data_lock):
lockfile_sleep(data_lock)
# create data lock file
create_lock_file(data_lock)
recognition_dict = get_json_data(recognition_json_path)
# delete data lock file
delete_lock_file(data_lock)
# name_dict : use product_id to get product_name
name_dict = {}
path = CUSTOMER_FILE + USERFILE
files = listdir(path)
for file in files:
if isdir(path + '/' + file):
data_data = get_json_data(path + '/' + file + '/data.data')
product_name = data_data["product_name"]
product_id = data_data["product_id"]
name_dict[product_id] = product_name
total_percentage = 0
image_count = len(response_data["items"])
items_datas = {}
htmls = []
data_locations = []
for i in range(image_count):
item_datas = {}
# set original data-key, anchorWidth and anchorHeight to response
response_data["items"][i]["id"] = keep_data[i][0]
response_data["items"][i]["AnchorWidth"] = keep_data[i][1]
response_data["items"][i]["AnchorHeight"] = keep_data[i][2]
response_data["items"][i]["b64string"] = PREBASE64 + response_data[
"items"][i]["b64string"]
iamge_items_count = len(response_data["items"][i]["item"])
percentage_response = response_data["items"][i]["percentage"]
total_percentage += percentage_response
# turn model id and name to customer product id and name.
for l in range(iamge_items_count):
model_id = str(response_data["items"][i]["item"][l]["real_id"])
max_id = ''
if model_id in recognition_dict:
max_count = 0
for key in recognition_dict[model_id].keys():
if recognition_dict[model_id][key] > max_count:
max_count = recognition_dict[model_id][key]
max_id = key
if max_id == '':
max_id = random_product_id(username)
response_data["items"][i]["item"][l]["real_id"] = max_id
max_name = name_dict[max_id]
if max_id in item_datas:
item_datas[max_id]["count"] += 1
item_datas[max_id]["position"].append(
response_data["items"][i]["item"][l]["position"])
else:
item_datas[max_id] = {
"AnchorWidth":
keep_data[i][1],
"AnchorHeight":
keep_data[i][2],
"count":
1,
"name":
max_name,
"position":
[response_data["items"][i]["item"][l]["position"]],
"cam_id":
response_data["items"][i]["id"],
"real_id":
max_id
}
if max_id in items_datas:
items_datas[max_id]["count"] += 1
else:
items_datas[max_id] = {"count": 1, "name": max_name}
# item_datas is one image all product id and name and count
html = html_camera_roi_inference_table_maker(
item_datas, percentage_response, False)
data_locations.append(item_datas)
htmls.append({"key": keep_data[i][0], "html": html})
total_percentage = total_percentage / image_count
html = html_camera_roi_inference_table_maker(items_datas,
total_percentage, True)
htmls.append({"key": "total", "html": html})
response_data["html"] = htmls
response_data["position"] = data_locations
return response_data