Spaces:
Sleeping
title: SimilarityThresholding
emoji: 💻
colorFrom: yellow
colorTo: yellow
sdk: gradio
sdk_version: 3.24.1
app_file: app.py
pinned: false
Zero-shot Segmentation Demo
该项目包含 zero-shot segmentation的两种用例。并且提供本地运行版本,以及在线版本。
Segmentation
Instance Retrieval
本地版本
python demos/retrive.py --pretrained_weights checkpoint.pth --k_img imgs/sample5.JPEG --q_img imgs/mix.jpg
python main.py --img imgs/mix.jpg --pretrained_weights checkpoint.pth
在线应用
Image Matching API
Overview
请注意,这个示例仅用于说明API接口的基本结构。实际实现时,您需要根据您的需求和技术选择来完成图像匹配和其他相关功能。
Person:
- 请设计详细的API接口文档并用markdown格式回答
ChatGPT:
- 以下是使用Markdown格式编写的详细API接口文档:
Image Matching API
本API提供图像匹配服务,允许用户提交两张图片("key" 和 "query"),并根据用户在"query"图像上选择的区域,返回匹配的物体。
1. Submit
Endpoint : /submit
Method : POST
Description : 用户在前端上传两张图片("key" 和 "query"),服务器收到后将图片保存在云端并生成唯一的image id。返回image id。
Request
- Content-Type :
multipart/form-data
ParameterTypeDescriptionkey_imageFile用户上传的"key"图片query_imageFile用户上传的"query"图片
Response
- Content-Type :
application/json
ParameterTypeDescriptionkey_image_idString"key"图片的唯一IDquery_image_idString"query"图片的唯一ID
Example
Request
POST /submit HTTP/1.1
Content-Type: multipart/form-data; boundary=----WebKitFormBoundaryyrV7KO0GIxCAQejs
------WebKitFormBoundaryyrV7KO0GIxCAQejs
Content-Disposition: form-data; name="key_image"; filename="key_image.jpg"
Content-Type: image/jpeg
(binary content)
------WebKitFormBoundaryyrV7KO0GIxCAQejs
Content-Disposition: form-data; name="query_image"; filename="query_image.jpg"
Content-Type: image/jpeg
(binary content)
------WebKitFormBoundaryyrV7KO0GIxCAQejs--
Response
{
"key_image_id": "550e8400-e29b-41d4-a716-446655440000",
"query_image_id": "6ba7b810-9dad-11d1-80b4-00c04fd430c8"
}
3. Get Image
Endpoint : /get_image
Method : GET
Description : 前端根据image id获取对应的图片。
RequestParameterTypeDescriptionimage_idString图片的唯一ID
Response
- Content-Type :
image/jpeg
Example
Request
GET /get_image?image_id=550e8400-e29b-41d4-a716-446655440000 HTTP/1.1
Response
HTTP/1.1 200 OK
Content-Type: image/jpeg
(binary content)
2. match
Endpoint : /match
Method : POST
Description : 用户选择"query"图像中的某个区域后,触发查询请求。前端发送"key" 和 "query"的image id,以及点击图片的坐标位置给服务器。服务器返回一个列表包含所有匹配的图片。
Request
- Content-Type :
application/json
ParameterTypeDescriptionkey_image_idString"key"图片的唯一IDquery_image_idString"query"图片的唯一IDcoordsObject用户点击的坐标位置,包含"x"和"y"属性
Response
- Content-Type :
application/json
ParameterTypeDescriptionimage_idString匹配图片的唯一IDurlString匹配图片的URL
Example
Request
POST /select HTTP/1.1
Content-Type: application/json
{
"key_image_id": "550e8400-e29b-41d4-a716-446655440000",
"query_image_id": "6ba7b810-9
}
2. segment
Endpoint : /segment
Method : POST
Description : 用户选择"query"图像中的某个区域后,触发查询请求。前端发送"query"的image id,以及点击图片的坐标位置给服务器。服务器返回一个列表包含所有匹配的图片。
Request
- Content-Type :
application/json
ParameterTypeDescriptionkey_image_idString"key"图片的唯一IDquery_image_idString"query"图片的唯一IDcoordsObject用户点击的坐标位置,包含"x"和"y"属性
Response
- Content-Type :
application/json
ParameterTypeDescriptionimage_idString匹配图片的唯一IDurlString匹配图片的URL
Example
Request
POST /select HTTP/1.1
Content-Type: application/json
{
"key_image_id": "550e8400-e29b-41d4-a716-446655440000",
"query_image_id": "6ba7b810-9
}