使用OpenAi的食品识别和营养估算（估算.识别.营养.食品.OpenAi...）

这是您可以在短短20分钟内使用openai构建简单的食物识别和营养估算应用程序的方法 它的工作原理

>图像编码：图像被转换为​​base64格式，以通过openai的api处理。

>食物识别提示：该应用将图像发送到openai，以识别食物及其各自的数量。

营养估计：使用另一个提示来估计基于确定的食品及其数量的营养价值。

> 显示结果：使用gradio显示出估计的卡路里，蛋白质，脂肪和碳水化合物的值。

>

这是一个非常简单的代码，可以改进/更好地组织起来，但是想法是说明它可以轻松地创建一个简单的poc。 如果您正在从事有趣的项目，请在

上与我联系

from openai import OpenAI
from pydantic import BaseModel
import base64
from typing import List
import gradio as gr

def encode_image(image_path):
  with open(image_path, "rb") as image_file:
    return base64.b64encode(image_file.read()).decode('utf-8')

openai_api_key = "key"
client = OpenAI(api_key=openai_api_key)

"""pydantic models to record food items and nutrient information, 
not necessary but helpful if you intend to create apis 
or use the data in other ways.
"""
class Food(BaseModel):
    name: str
    quantity: str

class Items(BaseModel):
    items: List[Food]

class Nutrient(BaseModel):
    steps: List[str]
    reasons: str
    kcal: str
    fat: str
    proteins: str
    carbohydrates: str


def recognize_items(image):
    """This function takes an image and returns a list of recognized food items along with their count and the nutrition. 
    """
    #first recognize items and quantities
    messages = [
        {
        "role": "user",
        "content": [
            {
            "type": "text",
            "text": f"You are an expert in recognising individual food items and their quantity. Give count(number) for countable items and an estimate for liquid/mixed or non countable items.  For example if you have one burger,two pastries, 2 pav, bhaji and dal in an image, you return burger,pastry,pav, bhaji and dal along with the count or estimates without any duplicates. For non countable items give an estimate in grams while explaining like 'looks 1 teaspoon of sauce, so around 5-8 grams' or 'looks 1 serving of bhaji, so around 150-200gms'. Given the image below, recognise food items with their quantity.",
            }
        ],
        }
    ]

    base64_image = encode_image(image)
    dic = {
                "type": "image_url",
                "image_url": {
                    "url":  f"data:image/jpeg;base64,{base64_image}",
                    "detail": "low"
                },
            }
    messages[0]["content"].append(dic)
    response = client.beta.chat.completions.parse(
    model="gpt-4o-mini",
    messages=messages,
    response_format=Items,
    max_tokens=300,
    temperature=0.1
    )
    foods = response.choices[0].message.parsed

    res = ""
    for food in foods.items:
        res=res+food.name+ " "+food.quantity+"
"

    #now estimate nutrition, we can use a separate model for this task
    messages = [
        {
        "role": "user",
        "content": [
            {
            "type": "text",
            "text": f"You are an expert in estimating information regarding nutririon given the food items and thier quantities. Think step by step considering the given food items and their quantities, and give an estimated range(lowest - highest) of kcal, range(lowest - highest) of fat, range of proteins(lowest - highest) and carbohydrates(lowest - highest). Ignore contributions from minor items. Ensure your estimations are solely based on the provided quantities.  Return steps,reasons and estimations if this food was consumed. 

food and quantity consumed by user: {res} 

.",
            }
        ],
        }
    ]
    dic = {
                "type": "image_url",
                "image_url": {
                    "url":  f"data:image/jpeg;base64,{base64_image}",
                    "detail": "low"
                },
            }
    messages[0]["content"].append(dic)
    response = client.beta.chat.completions.parse(
    model="gpt-4o-mini",
    messages=messages,
    response_format=Nutrient,
    max_tokens=500,
    temperature=0.1
    )
    nuts = response.choices[0].message.parsed
    steps = " ".join(nuts.steps)
    res=res+"
"+steps+"

calories: "+nuts.kcal+" 
fats: "+nuts.fat+" 
proteins: "+nuts.proteins+" 
carbohydrates: "+nuts.carbohydrates+"
"+nuts.reasons+"
"+"*These are estimations based on image. They might not be perfect or accurate. Please calculate based on the food you consume for a more precise estimate."
    return res


with gr.Blocks() as demo:
    foods=None
    with gr.Row():
        image_input = gr.Image(label="Upload Image",height=300,width=300,type="filepath")

    with gr.Row() as but_row:
        submit_btn = gr.Button("Detect food and quantity")

    with gr.Row() as text_responses_row: 
        text_response_1 = gr.Textbox(label="Detected food and quantity",scale=1)

    submit_btn.click(
        recognize_items,
        inputs=[image_input],
        outputs=[text_response_1]
    )

if __name__ == "__main__":
    demo.launch() 

以上就是使用OpenAi的食品识别和营养估算的详细内容，更多请关注知识资源分享宝库其它相关文章！