File size: 6,701 Bytes
754ac61 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 |
import requests
import json
import base64
import os
import time
import pandas as pd
from tqdm import tqdm
from concurrent.futures import ThreadPoolExecutor
DEBUG = True
SK_LIST = [
"sk-V8HaW6l4a6qkTRlR07423f3c8c67431c8a9d9c365c0b7d9b",
"sk-MZQXlv5tEG5hDX3yoK6sKRB4P9JBuw8PWtbeix1JITHWzIxW",
"sk-NgALyBkzs6LPt5kbvLS8WBILov33pL2rB6J5bLTI4FBk7O2p",
"sk-MEuJz0u5CyFyVgEP9CvUPhybfkP9eQg8iak82OU9pN6GC0xH",
]
COCO_ROOT = '/root/autodl-tmp/data/location_bench1/coco_resize/' # 根据需要修改路径
# 将图片转base64格式
def encode_image(image_path):
with open(image_path, "rb") as image_file:
return base64.b64encode(image_file.read()).decode('utf-8')
def gemini_label(question, img,idx, sk, attempt=0):
if attempt > 5:
return None
# 构建图片路径并编码为base64
# image_path = os.path.join(COCO_ROOT, 'val{0:06}.jpg'.format(idx)) # 根据需要修改路径
# image_path = f'/home/aiops/wangzh/data/RGBD-benchmark/out_doors/pic_all/{img}'
image_path = img
# image_path = '/home/aiops/wangzh/data/RGBD-benchmark/out_doors/pic_all/1_gpt4o.png'
base64_image = encode_image(image_path)
# import pdb;pdb.set_trace()
url = "https://open.xiaojingai.com/v1/chat/completions" # 替换为新API的URL
# "text": f'''You will see an image along with four corresponding descriptions (captions). Please carefully observe the image and select the description that best matches the content of the image. Choose one option from (A), (B), (C), or (D).
# Options: (A){question[0]}\n(B){question[1]}\n(C){question[2]}\n(D){question[3]}\nPlease provide your answer with only one of the options and nothing else.'''
try:
# 构建请求体
# payload = json.dumps({
# "image": f"data:image/jpeg;base64,{base64_image}",
# "question": question
# })
payload = json.dumps({
"model": "gemini-1.5-pro",
# "model": "gpt-4o",
"stream": False,
"messages": [
{
"role": "user",
"content": [
{
"type": "text",
"text": f'''You will see an image along with four corresponding descriptions (captions). Please carefully observe the image and select the description that best matches the content of the image. Choose one option from (A), (B), (C), or (D).
Options: (A){question[0]}\n(B){question[1]}\n(C){question[2]}\n(D){question[3]}\nPlease provide your answer with only one of the options and nothing else.'''
},
{
"type": "image_url",
"image_url": {
"url": f"data:image/jpeg;base64,{base64_image}"
}
}
]
}
],
"max_tokens": 1000
})
headers = {
'Authorization': sk,
'Content-Type': 'application/json'
}
# 发送POST请求
response = requests.post(url, headers=headers, data=payload)
print(response)
print("Response Status Code:", response.status_code) # 打印状态码
print("Response Text:", response.text) # 打印响应内容
output = response.json() # 获取回答
print("output",output)
return {"id": idx, "answer": output}
except Exception as ex:
print(idx, ex)
time.sleep(2) # 等待后重试
return gemini_label(question, idx, sk, attempt + 1) # 重试
def process_sample(i, img,questions, sk):
uid = uids[i]
question = questions[i]
img = img[i]
# import pdb;pdb.set_trace()
ans = gemini_label(question, img,uid, sk=sk)
return ans if ans is not None else 'None'
if __name__ == '__main__':
save_path = './'
os.makedirs(os.path.join(save_path), exist_ok=True)
# meta_path = '/home/aiops/wangzh/data/RGBD-benchmark/out_doors/all.json'
meta_path = '/home/aiops/wangzh/data/scanner/indoor-new/all.json'
# meta = pd.read_csv(meta_path)
with open(meta_path, 'r', encoding='utf-8') as json_file:
meta = json.load(json_file)
import pdb;pdb.set_trace()
lens=len(meta)
uids = [meta[i]['id'] for i in range(lens)]
questions =[meta[i]['captions'] for i in range(lens)]
img = [meta[i]['image'] for i in range(lens)]
# print('get samples:', len(uids))
# print(uids[:5])
# img = [f'/home/aiops/wangzh/data/scanner/scannet_2d_HR3/{meta[i]['scene_id']}/color/{meta[i]['image']}' for i in range(lens)]
img = [f'/home/aiops/wangzh/data/scanner/scannet_2d_HR3/{meta[i]["scene_id"]}/color/{meta[i]["image"]}' for i in range(lens)]
answer_list = []
with ThreadPoolExecutor(max_workers=4) as executor:
answer_list = list(tqdm(executor.map(lambda i: process_sample(i, img,questions, SK_LIST[i % 4]), range(len(uids))), total=len(uids)))
print('gemini label sample:', len(answer_list))
answer_list = pd.DataFrame(answer_list)
answer_list.to_csv(os.path.join(save_path, 'answers.csv'), index=False)
# import requests
# import json
# import base64
# import os
# def encode_image(image_path):
# with open(image_path, "rb") as image_file:
# return base64.b64encode(image_file.read()).decode('utf-8')
# def ask_question(image_path, caption, sk):
# base64_image = encode_image(image_path)
# url = "https://open.xiaojingai.com/v1/chat/completions"
# payload = json.dumps({
# "model": "gemini-1.5-pro",
# "stream": False,
# "messages": [
# {
# "role": "user",
# "content": [
# {
# "type": "text",
# "text":"describe what is love"
# },
# {
# "type": "image_url",
# "image_url": {
# "url": f"data:image/jpeg;base64,{base64_image}"
# }
# }
# ]
# }
# ],
# "max_tokens": 1000
# })
# headers = {
# 'Authorization': sk,
# 'Content-Type': 'application/json'
# }
# response = requests.post(url, headers=headers, data=payload)
# return response.json()
# # Example usage
# image_path = '/home/aiops/wangzh/data/RGBD-benchmark/out_doors/pic_all/1_gpt4o.png'
# caption = "Describe the content of the image."
# sk = "sk-V8HaW6l4a6qkTRlR07423f3c8c67431c8a9d9c365c0b7d9b"
# response = ask_question(image_path, caption, sk)
# print(response) |