图片理解
Gemini 模型从一开始就具有多模态特性,可执行各种图像处理和计算机视觉任务,包括但不限于图片说明、分类和视觉问答,而无需训练专门的机器学习模型。
除了通用的多模态功能外,Gemini 模型还通过额外的训练,针对特定应用场景(例如对象检测和细分)提供更高的准确性。
向 Gemini 传递图片
您可以使用多种方法将图片作为输入内容提供给 Gemini:
- 使用网址传递图片:非常适合可公开访问的图片。
- 传递内嵌图片数据:用于传递 base64 编码的图片数据。
- 使用 File API 上传图片:建议用于较大的文件或在多个请求中重复使用图片。
使用网址传递图片
您可以使用 Files API 上传图片,并在请求中传递该图片:
Python
from google import genai
client = genai.Client()
uploaded_file = client.files.upload(file="path/to/organ.jpg")
interaction = client.interactions.create(
model="gemini-3-flash-preview",
input=[
{"type": "text", "text": "Caption this image."},
{
"type": "image",
"uri": uploaded_file.uri,
"mime_type": uploaded_file.mime_type
}
]
)
print(interaction.steps[-1].content[0].text)
JavaScript
import { GoogleGenAI } from "@google/genai";
const client = new GoogleGenAI({});
const uploadedFile = await client.files.upload({
file: "path/to/organ.jpg",
config: { mimeType: "image/jpeg" }
});
const interaction = await client.interactions.create({
model: "gemini-3-flash-preview",
input: [
{type: "text", text: "Caption this image."},
{
type: "image",
uri: uploadedFile.uri,
mimeType: uploadedFile.mimeType
}
]
});
console.log(interaction.steps.at(-1).content[0].text);
REST
# First upload the file using the Files API, then use the URI:
curl -X POST "https://generativelanguage.googleapis.com/v1beta/interactions" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H 'Content-Type: application/json' \
-d '{
"model": "gemini-3-flash-preview",
"input": [
{"type": "text", "text": "Caption this image."},
{
"type": "image",
"uri": "YOUR_FILE_URI",
"mime_type": "image/jpeg"
}
]
}'
传递内嵌图片数据
您可以以 base64 编码的字符串形式提供图片数据:
Python
from google import genai
with open('path/to/small-sample.jpg', 'rb') as f:
image_bytes = f.read()
client = genai.Client()
interaction = client.interactions.create(
model="gemini-3-flash-preview",
input=[
{"type": "text", "text": "Caption this image."},
{
"type": "image",
"data": base64.b64encode(image_bytes).decode('utf-8'),
"mime_type": "image/jpeg"
}
]
)
print(interaction.steps[-1].content[0].text)
JavaScript
import { GoogleGenAI } from "@google/genai";
import * as fs from "node:fs";
const client = new GoogleGenAI({});
const base64ImageFile = fs.readFileSync("path/to/small-sample.jpg", {
encoding: "base64",
});
const interaction = await client.interactions.create({
model: "gemini-3-flash-preview",
input: [
{type: "text", text: "Caption this image."},
{
type: "image",
data: base64ImageFile,
mime_type: "image/jpeg"
}
]
});
console.log(interaction.steps.at(-1).content[0].text);
REST
IMG_PATH="/path/to/your/image1.jpg"
if [[ "$(base64 --version 2>&1)" = *"FreeBSD"* ]]; then
B64FLAGS="--input"
else
B64FLAGS="-w0"
fi
curl -X POST "https://generativelanguage.googleapis.com/v1beta/interactions" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H 'Content-Type: application/json' \
-d '{
"model": "gemini-3-flash-preview",
"input": [
{"type": "text", "text": "Caption this image."},
{
"type": "image",
"data": "'"$(base64 $B64FLAGS $IMG_PATH)"'",
"mime_type": "image/jpeg"
}
]
}'
使用 File API 上传图片
对于大型文件,或者为了能够重复使用同一图片文件,请使用 Files API。请参阅 Files API 指南。
Python
from google import genai
client = genai.Client()
my_file = client.files.upload(file="path/to/sample.jpg")
interaction = client.interactions.create(
model="gemini-3-flash-preview",
input=[
{"type": "text", "text": "Caption this image."},
{
"type": "image",
"uri": my_file.uri,
"mime_type": my_file.mime_type
}
]
)
print(interaction.steps[-1].content[0].text)
JavaScript
import { GoogleGenAI } from "@google/genai";
const client = new GoogleGenAI({});
const myfile = await client.files.upload({
file: "path/to/sample.jpg",
config: { mimeType: "image/jpeg" },
});
const interaction = await client.interactions.create({
model: "gemini-3-flash-preview",
input: [
{type: "text", text: "Caption this image."},
{
type: "image",
uri: myfile.uri,
mime_type: myfile.mimeType
}
]
});
console.log(interaction.steps.at(-1).content[0].text);
REST
# First upload the file (see Files API guide for details)
# Then use the file URI in the request:
curl -X POST "https://generativelanguage.googleapis.com/v1beta/interactions" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H 'Content-Type: application/json' \
-d '{
"model": "gemini-3-flash-preview",
"input": [
{"type": "text", "text": "Caption this image."},
{
"type": "image",
"uri": "YOUR_FILE_URI",
"mime_type": "image/jpeg"
}
]
}'
使用多张图片进行提示
您可以在单个提示中提供多张图片,只需在 input 数组中添加多个图片对象即可:
Python
from google import genai
client = genai.Client()
interaction = client.interactions.create(
model="gemini-3-flash-preview",
input=[
{"type": "text", "text": "What is different between these two images?"},
{
"type": "image",
"uri": "https://example.com/image1.jpg",
"mime_type": "image/jpeg"
},
{
"type": "image",
"uri": "https://example.com/image2.jpg",
"mime_type": "image/jpeg"
}
]
)
print(interaction.steps[-1].content[0].text)
JavaScript
import { GoogleGenAI } from "@google/genai";
const client = new GoogleGenAI({});
const interaction = await client.interactions.create({
model: "gemini-3-flash-preview",
input: [
{type: "text", text: "What is different between these two images?"},
{
type: "image",
uri: "https://example.com/image1.jpg",
mime_type: "image/jpeg"
},
{
type: "image",
uri: "https://example.com/image2.jpg",
mime_type: "image/jpeg"
}
]
});
console.log(interaction.steps.at(-1).content[0].text);
REST
curl -X POST "https://generativelanguage.googleapis.com/v1beta/interactions" \
-H "x-goog-api-key: $GEMINI_API_KEY" \
-H 'Content-Type: application/json' \
-d '{
"model": "gemini-3-flash-preview",
"input": [
{"type": "text", "text": "What is different between these two images?"},
{
"type": "image",
"uri": "https://example.com/image1.jpg",
"mime_type": "image/jpeg"
},
{
"type": "image",
"uri": "https://example.com/image2.jpg",
"mime_type": "image/jpeg"
}
]
}'
对象检测
模型经过训练,可检测图片中的对象并获取其边界框坐标。相对于图片尺寸的坐标会缩放到 [0, 1000]。您需要根据原始图片大小对这些坐标进行反缩放。
Python
from google import genai
from PIL import Image
import json
client = genai.Client()
prompt = "Detect the all of the prominent items in the image. The box_2d should be [ymin, xmin, ymax, xmax] normalized to 0-1000."
interaction = client.interactions.create(
model="gemini-3-flash-preview",
input=[
{"type": "text", "text": prompt},
{
"type": "image",
"uri": "https://example.com/image.png",
"mime_type": "image/png"
}
],
response_format={
"type": "text",
"mime_type": "application/json"
}
)
bounding_boxes = json.loads(interaction.steps[-1].content[0].text)
print("Bounding boxes:", bounding_boxes)
如需查看更多示例,请参阅 Gemini 实战宝典中的以下笔记本:
细分
从 Gemini 2.5 开始,模型不仅可以检测项目,还可以对项目进行分割并提供其轮廓遮罩。
模型会预测一个 JSON 列表,其中每个项都表示一个分割掩码。每个商品都有一个边界框 (“box_2d”),格式为 [y0, x0, y1, x1],其中包含介于 0 到 1000 之间的归一化坐标;一个用于标识对象的标签 (“label”);最后是边界框内的分割掩码,以 base64 编码的 PNG 格式表示,是一个介于 0 到 255 之间的概率图。
Python
from google import genai
from PIL import Image
import json
client = genai.Client()
prompt = """
Give the segmentation masks for the wooden and glass items.
Output a JSON list of segmentation masks where each entry contains the 2D
bounding box in the key "box_2d", the segmentation mask in key "mask", and
the text label in the key "label". Use descriptive labels.
"""
interaction = client.interactions.create(
model="gemini-3-flash-preview",
input=[
{"type": "text", "text": prompt},
{
"type": "image",
"uri": "https://example.com/image.png",
"mime_type": "image/png"
}
],
config={
"thinking_level": "minimal" # Minimize thinking for better detection results
}
)
items = json.loads(interaction.steps[-1].content[0].text)
print("Segmentation results:", items)
支持的图片格式
Gemini 支持以下图片格式 MIME 类型:
- PNG -
image/png - JPEG -
image/jpeg - WEBP -
image/webp - HEIC -
image/heic - HEIF -
image/heif
如需了解其他文件输入方法,请参阅文件输入方法指南。
功能
所有 Gemini 模型版本都是多模态模型,可用于各种图像处理和计算机视觉任务,包括但不限于图片说明、视觉问答、图片分类、对象检测和分割。
Gemini 可以减少对专用机器学习模型的需求,具体取决于您的质量和性能要求。
最新版本的模型经过专门训练,除了通用功能外,还可提高专业化任务的准确率,例如增强的对象检测和细分。
限制和关键技术信息
文件数量限制
Gemini 模型支持每个请求最多上传 3,600 个图片文件。
token 计算
- 如果两个维度均小于或等于 384 像素,则为 258 个 token。 较大的图片会被分块为 768x768 像素的图块,每个图块需花费 258 个 token。
计算图块数量的粗略公式如下:
- 计算裁剪单元大小(大致为:
floor(min(width, height)/ 1.5)。 - 将每个维度除以裁剪单元大小,然后将结果相乘,即可得到图块数量。
例如,对于尺寸为 960x540 的图片,剪裁单元尺寸为 360。将每个维度除以 360,得到的图块数量为 3 * 2 = 6。
媒体分辨率
Gemini 3 引入了 media_resolution 参数,可对多模态视觉处理进行精细控制。media_resolution 参数用于确定为每个输入图片或视频帧分配的 token 数量上限。分辨率越高,模型读取细小文字或识别细微细节的能力就越强,但 token 用量和延迟时间也会增加。
技巧和最佳做法
- 验证图片是否已正确旋转。
- 使用清晰且不模糊的图片。
- 如果使用包含文本的单张图片,请在
input数组中将文本提示放在图片之前。
后续步骤
本指南将介绍如何上传图片文件并根据图片输入生成文本输出。如需了解详情,请参阅以下资源: