jiahui
/
ai_creative_workshop
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity
Browse Source

ai创意工厂需要的接口组合

master
majiahui@haimaqingfan.com 2 years ago
parent
3630743174
commit
062b5f7daf
1 changed files with 185 additions and 90 deletions
Whitespace
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Split View
Diff Options
Show Stats Download Patch File Download Diff File
  1. 197
      main.py

197
main.py
View File

@ -15,9 +15,11 @@ app.config['UPLOAD_FOLDER'] = UPLOAD_FOLDER
RE_CHINA_NUMS = "[1-9].(.*)"
# 允许的文件类型
ALLOWED_EXTENSIONS = {'png', 'jpg', 'jpeg', 'gif'}
fenhao_list = [";", "；"]
prompt_picture = {
prompt_picture_dict = {
"1": "图中的商品：{}，有什么突出亮点和卖点，请分条列举出来，要求亮点或者卖点要用一个词总结，冒号后面在进行解释，例如：1. 时尚黑色：图中的鞋子是黑色的，符合时尚潮流，适合不同场合的穿搭。",
"2": "图中的商品：{}，有什么亮点，写一段营销话语",
"3": "图中的商品：{}，有以下亮点：\n{}\n根据这些优势亮点，写一段营销文本让商品卖的更好",
@ -26,14 +28,13 @@ prompt_picture = {
"6": "根据图中的商品：{}，生成一个商品名称，要求商品名称格式中包含的信息（有品牌名，有产品名，有细分产品种类词，比如猫砂，篮球鞋等，有三到五个卖点和形容词）",
}
# prompt_text = {
# "1": "图中{}有什么突出亮点，请列举出来",
# "2": "图中{}有什么亮点，写一段营销话语",
# "3": "图中{}有以下亮点：\n{}\n根据这些优势亮点，写一段营销文本让商品买的更好",
# "4": "图中{}有哪些不足之处可以改进？",
# "5": "图中{}的渲染图做哪些调整可以更吸引消费者",
# "5": "图中{}的渲染图做哪些调整可以更吸引消费者",
# }
prompt_text_dict = {
"1": "",
"2": "User:商品名称:{};卖点:{}，请帮我生成一个有很多活泼表情的小红书文案，以商品使用者角度来写作，让人感觉真实\nAssistant:",
"3": "图中{}有以下亮点：\n{}\n根据这些优势亮点，写一段营销文本让商品买的更好",
"4": "图中{}有哪些不足之处可以改进？",
"5": "图中{}的渲染图做哪些调整可以更吸引消费者",
}
def dialog_line_parse(url, text):
@ -88,7 +89,7 @@ def allowed_file(filename):
return '.' in filename and filename.rsplit('.', 1)[1].lower() in ALLOWED_EXTENSIONS
def picyure_model_predict(image_path, prompt):
def picture_model_predict(image_path, prompt):
# query = tokenizer.from_list_format([
# {'image': image_path},
# {'text': prompt},
@ -106,16 +107,35 @@ def picyure_model_predict(image_path, prompt):
return result
def picture_main(path_list, commodity, type, additional):
if type == "1":
def text_model_predict(prompt):
# query = tokenizer.from_list_format([
# {'image': image_path},
# {'text': prompt},
# ])
#
# response, history = model.chat(tokenizer, query=query, history=None)
# return response
url = "http://192.168.31.74:12000/predict"
data = {
"texts": prompt,
}
result = dialog_line_parse(url, data)["data"]
return result
def type_1(path_list, commodity, input_type):
code = 200
result_list_len = False
dan_result_geshi = True
dan_result_geshi_maohao = True
prompy_text = prompt_picture[type]
return_list = []
prompy_text = prompt_picture_dict[input_type]
prompy_text = prompy_text.format(commodity)
for path in path_list:
while True:
result = picyure_model_predict(path_list[0], prompy_text)
result = picture_model_predict(path, prompy_text)
result_list = str(result).split("\n")
result_list = [i for i in result_list if i != ""]
if len(result_list) > 3:
@ -137,39 +157,86 @@ def picture_main(path_list, commodity, type, additional):
guanjianci = response_re[0].split("：")
maidian_list.append([i, guanjianci])
return maidian_list
return_list.append(maidian_list)
return code, return_list
elif type == "2":
prompy_text = prompt_picture[type]
prompy_text = prompy_text.format(commodity, additional)
def type_2(path_list, commodity, input_type, additional):
code = 200
return_list = []
for path in path_list:
result = picyure_model_predict(path, prompy_text)
return_1_data = type_1(path_list, commodity, "1")
maidian = [i[0][1][0] for i in return_1_data]
fenhao = ""
if additional != "":
for i in fenhao_list:
if i in additional:
fenhao = i
break
if fenhao == "":
return code, []
maidian_user = [i for i in additional.split(fenhao) if i != ""]
maidian += maidian_user
prompt_text = prompt_text_dict[input_type].format(commodity, "、".join(maidian))
result = text_model_predict(prompt_text)
return_list.append(result)
return return_list
#
elif type == "3":
prompy_text = prompt_picture[type]
prompy_text = prompy_text.format(commodity, additional)
return code, return_list
def type_3(path_list, commodity, input_type, additional):
code = 200
return_list = []
for path in path_list:
result = picyure_model_predict(path, prompy_text)
return_1_data = type_1(path_list, commodity, "1")
maidian = [i[0][1][0] for i in return_1_data]
fenhao = ""
if additional != "":
for i in fenhao_list:
if i in additional:
fenhao = i
break
if fenhao == "":
return code, []
maidian_user = [i for i in additional.split(fenhao) if i != ""]
maidian += maidian_user
prompt_text = prompt_text_dict[input_type].format(commodity, "、".join(maidian))
result = text_model_predict(prompt_text)
return_list.append(result)
return return_list
#
elif type == "4":
prompy_text = prompt_picture[type]
prompy_text = prompy_text.format(commodity, additional)
return code, return_list
def type_4(path_list, commodity, input_type, additional):
code = 200
return_list = []
for path in path_list:
for i in range(5):
result = picyure_model_predict(path, prompy_text)
return_1_data = type_1(path_list, commodity, "1")
maidian = [i[0][1][0] for i in return_1_data]
fenhao = ""
if additional != "":
for i in fenhao_list:
if i in additional:
fenhao = i
break
if fenhao == "":
return code, []
maidian_user = [i for i in additional.split(fenhao) if i != ""]
maidian += maidian_user
prompt_text = prompt_text_dict[input_type].format(commodity, "、".join(maidian))
result = text_model_predict(prompt_text)
return_list.append(result)
return return_list
return code, return_list
elif type == "5":
def type_5(path_list, commodity, input_type):
code = 200
return_list = []
prompy_text = prompt_picture[type]
prompy_text = prompt_picture_dict[input_type]
prompy_text = prompy_text.format(commodity)
result_list_type = False
@ -177,7 +244,7 @@ def picture_main(path_list, commodity, type, additional):
while True:
if result_list_type == True:
break
result = picyure_model_predict(path, prompy_text)
result = picture_model_predict(path, prompy_text)
result_list = str(result).split("\n")
result_list = [i for i in result_list if i != ""]
result_list_new = []
@ -191,25 +258,53 @@ def picture_main(path_list, commodity, type, additional):
result_list_type = True
return_list.append(result_list_new)
return return_list
return code, return_list
elif type == "6":
def type_6(path_list, commodity, input_type):
code = 200
return_list = []
prompy_text = prompt_picture[type]
commodity_list = []
prompy_text = prompt_picture_dict[input_type]
prompy_text = prompy_text.format(commodity)
for path in path_list:
for i in range(5):
result = picyure_model_predict(path, prompy_text)
return_list.append(result)
return return_list
result = picture_model_predict(path, prompy_text)
commodity_list.append(result)
return_list.append(commodity_list)
return code, return_list
elif type == "7":
def type_7(path_list, additional):
code = 200
prompy_text = additional
return_list = []
for path in path_list:
result = picyure_model_predict(path, prompy_text)
result = picture_model_predict(path, prompy_text)
return_list.append(result)
return return_list
return code, return_list
def picture_main(path_list, commodity, input_type, additional):
if input_type == "1":
return type_1(path_list, commodity, input_type)
elif input_type == "2":
return type_2(path_list, commodity, input_type, additional)
#
elif input_type == "3":
return type_3(path_list, commodity, input_type, additional)
#
elif input_type == "4":
return type_4(path_list, commodity, input_type, additional)
elif input_type == "5":
return type_5(path_list, commodity, input_type)
elif input_type == "6":
return type_6(path_list, commodity, input_type)
elif input_type == "7":
return type_7(path_list, additional)
else:
return "1111"
@ -267,9 +362,9 @@ def upload_file():
result = []
for type_dan in type_list:
print("type:", type_dan)
result_dan = picture_main(path_list, commodity, type_dan, additional)
code, result_dan = picture_main(path_list, commodity, type_dan, additional)
result.append(result_dan)
return_text = {"texts": result, "probabilities": None, "status_code": 200}
return_text = {"texts": result, "probabilities": None, "status_code": code}
except:
return_text = {"texts": "运算出错", "probabilities": None, "status_code": 400}
log.log('start at',

Write Preview
Loading…
Cancel
Save