Skip to content

Latest commit

 

History

History
1451 lines (1116 loc) · 42.8 KB

miaohua_task_submit.md

File metadata and controls

1451 lines (1116 loc) · 42.8 KB

注意事项

Warning

🔴❗Warning❗🔴 生图任务产生的图片，请及时自行下载保存，不保证链接时效，2023年10月18日之前生成的图片链接将在2023年11月初失效，可以通过任务结果接口重新获取新的图片链接。 [!WARNING]
🔴❗Warning❗🔴 生图任务调用模型请改用versionid字段( 可通过/api/v1b/models/base获得对应versionid)。❗❗将在2024年05月删除对model_name字段支持❗❗ [!WARNING]
🔴❗Warning❗🔴 调用模型训练请改用base_model_version字段❗❗将在2024年05月删除对model_name字段支持❗❗ [!WARNING]
🔴❗Warning❗🔴 图生文已经下线，不支持调用

常见问题解答

图片生成(i2i,t2i) 接口

获取token

请求地址

POST https://miaohua.sensetime.com/api/v1b/get_token

请求参数

参数名称 类型 是否必须 默认值 含义
email string 是 "[email protected]
" tob用户的用户邮箱，需先在miaohua.sensetime.com注册
password string 是 "password" tob用户密码

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/get_token'
data = {
   "email": "[email protected]",
   "password": "password",
}

response = requests.post(url, json=data)

print(response.status_code)
print(response.text)

返回示例

{
    "token": "xxxxxxxxxxxxxxxxxx", # string 当前用户token
}

任务提交

请求地址

POST https://miaohua.sensetime.com/api/v1b/task_submit

请求参数

参数名称 类型 是否必须 默认值 含义
token string 是 无，通过get_token获得 通过get_token获取的token
versionid string 是 sgl_artist_v0.3.5_0925 新增字段，模型版本id( 可通过/api/v1b/models/base获得对应versionid) versionid必须提交
prompt string 是 "" 用于生成图片的特征描述，如："one girl,beautiful"
neg_prompt string 否 "" 特征的反向描述，一般无需指定
n_images int 是 2 生成图片数量
scale int 是 7 文本控制力度(1-20), 同Prompt weight
strength float 是 0.6 图片控制力度(0-1), 同Image weight
ddim_steps int 否 50 迭代步数
select_seed int 否 -1 随机数种子
output_size string 是 960x960 图片的输出尺寸，如："960x960",长度范围可选640-6240，宽度范围可选640-6240
callback_url string 否 "" 回调url，url形式，对任务进度进行主动通知
init_img string 否 "" 输入图片，url形式，若有即为i2i(图生图)，无即为t2i(文生图)。必须是通过/api/v1b/upload_img或 /api/v1b/upload_imgs 上传后的图片地址
add_prompt bool 否 false 是否使用gpt进行描述词优化, 开启优化生图速度会有一定影响
cn_configs array 否 null controlnet配置，包括权重、图片，见示例. 支持的cn从 /api/v1b/cn/artist 或 /api/v1b/cn/opensource中获取. 权重范围为 0~2, 图片为通过upload_img/upload_imgs接口上传的图片地址，最多支持传入3个cn
lora_configs array 否 null lora配置， 包括lora模型、权重， 见示例. versionid从 /api/v1b/models/private 或 /api/v1b/models/lora中获取. 权重范围为 0~2
outpaint_config dict 否 null 扩图配置， 见示例{"src_img": url,"outpaint_ratio": 1.5,"align_position": "middle"}. src_img为上传的img url，outpaint_ratio取值范围是[1.5, 1.25, 2], align_position为topleft，topmiddle，topright，middleleft，middle，middleright，bottomleft，bottommiddle，bottomright

注意：

  1. 开源和自研模型的controlnet支持不一样，可以通过 /api/v1b/cn/artist 和 /api/v1b/cn/opensource获取自研和开源模型支持的cn
  2. 只有基模型才能作为初始模型，可以通过 /api/v1b/models/base获取基模型， 目前支持一层lora，多层不生效
  3. 设置outpaint_config无需提供versionid，versionid设置无效

请求示例

curl 示例

curl https://miaohua.sensetime.com/api/v1b/task_submit \
  -H "Content-Type: application/json" \
  -d '{
    "versionid": "sgl_artist_v0.3.5_0925",
    "prompt": "one girl, beautiful",
    "neg_prompt": "",
    "n_images": 2,
    "scale": 7,
    "output_size": "960x960",
    "select_seed": -1,
    "token": "",
    "init_img": "",
    "ddim_steps": 7,
    "strength": 0.82,
    "lora_configs": [
      {"versionid": "", "merge_weight": 1}
    ],
    "cn_configs": {
      {"cn": "openpose", "img": "https://ossxxxxxx.png", "weight": 1}
    }
  }'

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/task_submit'
data = {
    "versionid": "sgl_artist_v0.3.5_0925", # string 用到的模型versionid（规定范围内）
    "prompt": "one girl, beautiful", # 正向描述词
    "neg_prompt": "", # 反向描述词
    "n_images": 2, # int 生成图片的数量
    "ddim_steps": 50, # int 迭代步数
    "scale": 7, # int 文本控制力度
    "strength": 0.6, # float 图片控制力度
    "output_size": "960x960", # string 生成的图片尺寸
    "select_seed": -1, # int 随机数种子。如果是-1的话代表不指定
    "init_img": "", # img url,若为空为文生图，否则为图生图
    "add_prompt": False, # 是否使用gpt进行描述词优化
    "token": "",  # get_token获取的token
    "lora_configs": [
      {"versionid": "aaa", "merge_weight": 1}  # versionid从 /api/v1b/models/private 或 /api/v1b/models/lora中获取. 权重范围为 0~2
    ],
    "cn_configs": [
      {"cn": "openpose", "img": "https://ossxxxxxx.png", "weight": 1}  # 支持的cn从 /api/v1b/cn/artist 或 /api/v1b/cn/opensource中获取. 权重范围为 0~2, 图片为通过upload_img/upload_imgs接口上传的图片地址
    ]
}

response = requests.post(url, json=data)

print(response.status_code)
print(response.text)

controlnet 参数示例

{
        "cn": "tile", //['canny','lineart',..]
        "weight": 0.3,
        "guidance_start": 0.1, //  0-1 默认0
        "guidance_end": 0.9, //  0-1 默认1
        "img": "https://bkmk.oss-accelerate.aliyuncs.com/35c82400-52c8-11ee-817a-1e719ed4d673.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317054673034&Signature=X1A%2Bhnro0l0TmLOWKknlBckJhok%3D",
        "config": {//config可选，针对特定cn支持配置
            "control_processor": "tile_resample",
            "down_sample_rate": 1 // , tile需要 浮点数两位小数,默认1,范围1-8
        }
}

以下为cn现在支持的内置config 参数, 存在差异，按照文档调用

{
        "cn": "canny",
        "weight": 0.3,
        "guidance_start": 0.1, //  0-1 默认0
        "guidance_end": 0.9, //  0-1 默认1
        "img": "https://bkmk.oss-accelerate.aliyuncs.com/35c82400-52c8-11ee-817a-1e719ed4d673.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317054673034&Signature=X1A%2Bhnro0l0TmLOWKknlBckJhok%3D",
        "config": {
            "control_processor": "canny",//canny invert
            "low_threshold":1,// 1-255,
            "high_threshold": 255,// 1-255,
        }
}
{
        "cn": "lineart",
        "weight": 0.3,
        "guidance_start": 0.1, //  0-1 默认0
        "guidance_end": 0.9, //  0-1 默认1
        "img": "https://bkmk.oss-accelerate.aliyuncs.com/35c82400-52c8-11ee-817a-1e719ed4d673.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317054673034&Signature=X1A%2Bhnro0l0TmLOWKknlBckJhok%3D",
        "config": {
            "control_processor": "lineart_realistic",//'lineart_realistic','lineart_coarse','lineart_standard','invert'
        }
}
{
        "cn": "lineart_anime",
        "weight": 0.3,
        "guidance_start": 0.1, //  0-1 默认0
        "guidance_end": 0.9, //  0-1 默认1
        "img": "https://bkmk.oss-accelerate.aliyuncs.com/35c82400-52c8-11ee-817a-1e719ed4d673.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317054673034&Signature=X1A%2Bhnro0l0TmLOWKknlBckJhok%3D",
        "config": {
            "control_processor": "lineart_anime" or "lineart_anime_denoise",
        }
}
{
        "cn": "openpose",
        "weight": 0.3,
        "guidance_start": 0.1, //  0-1 默认0
        "guidance_end": 0.9, //  0-1 默认1
        "img": "https://bkmk.oss-accelerate.aliyuncs.com/35c82400-52c8-11ee-817a-1e719ed4d673.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317054673034&Signature=X1A%2Bhnro0l0TmLOWKknlBckJhok%3D",
        "config": {
            "control_processor": "dw_openpose_full,
            /*['dw_openpose_full',
                'openpose',
                'openpose_face',
                'openpose_faceonly',
                'openpose_full',
                'openpose_hand',]*/
        }
}
{
        "cn": "tile",
        "weight": 0.3,
        "guidance_start": 0.1, //  0-1 默认0
        "guidance_end": 0.9, //  0-1 默认1
        "img": "https://bkmk.oss-accelerate.aliyuncs.com/35c82400-52c8-11ee-817a-1e719ed4d673.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317054673034&Signature=X1A%2Bhnro0l0TmLOWKknlBckJhok%3D",
        "config": {
            "control_processor": "tile_colorfix,
            /*['tile_colorfix',
              'tile_colorfix_sharp',
              'tile_resample',]*/
            "down_sample_rate":1,//1-8

        }
}

多cn 示例

{
    "token": "",
    "lora_configs": [],
    "model_name": "Anything V3_fp16_train",
    "prompt": "t1#girl, room, red hair, red clothes",
    "neg_prompt": "",
    "n_images": 2,
    "scale": 7,
    "select_seed": -1,
    "init_img": "https://bkmk.oss-accelerate.aliyuncs.com/2c0001d0-4004-11ee-b30c-00163e005161.jpg?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317052609716&Signature=ZalLV8FSCsTU3DjPtKCqOn2qXr0%3D",
    "add_prompt": false,
    "ddim_steps": 50,
    "output_size": "2748x4096",
    "cn_configs": [{
        "cn": "tile",
        "weight": 0.3,
        "guidance_start": 0.1, //  0-1 默认0
        "guidance_end": 0.9, //  0-1 默认1
        "img": "https://bkmk.oss-accelerate.aliyuncs.com/35c82400-52c8-11ee-817a-1e719ed4d673.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317054673034&Signature=X1A%2Bhnro0l0TmLOWKknlBckJhok%3D",
        "config": {
            "control_processor": "tile_resample",
            "down_sample_rate": 1 // , tile需要 浮点数两位小数,默认1,范围1-8
        }
    }, {
        "cn": "brightness",
        "weight": 0.4,
        "guidance_start": 0.2, //  0-1 默认0
        "guidance_end": 0.8, //  0-1 默认1
        "img": "https://bkmk.oss-accelerate.aliyuncs.com/35c82400-52c8-11ee-817a-1e719ed4d673.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317054673034&Signature=X1A%2Bhnro0l0TmLOWKknlBckJhok%3D",
      
    }]
}
{
    "token": "",
    "lora_configs": [],
    "model_name": "Dalcefo",
    "prompt": "t1#girl, room, red hair, red clothes",
    "neg_prompt": "",
    "n_images": 2,
    "scale": 7,
    "select_seed": -1,
    "init_img": "https://bkmk.oss-accelerate.aliyuncs.com/8804c8b2-52c9-11ee-aa0d-ea71f9947415.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317054673602&Signature=hguT%2BR3VxxSvZLxBg%2FMi6vanyA8%3D",
    "add_prompt": false,
    "ddim_steps": 50,
    // "output_size": "2748x4096",
    "output_size": "2256x2252",
    "cn_configs": [{
        "cn": "canny",
        "weight": 0.9,
        "guidance_start": 0, //  0-1 默认0
        "guidance_end": 1, //  0-1 默认1
        "img": "https://bkmk.oss-accelerate.aliyuncs.com/aaa2d68e-52c4-11ee-817a-1e719ed4d673.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317054671512&Signature=hUkZUXxJ1eHCwgQ4m0UiWz%2BaC6o%3D",
        "config": {
            "control_processor": "canny",
            "low_threshold": 100, // , canny需要 范围1-255，默认100
            "high_threshold": 200 // , canny需要 范围1-255，默认200
        }
    	}]
}

返回示例

{
    "task_id": "f3e5b59c-7416-11ed-a160-00163e025c94", # string 任务id
}

outpaint 示例

 {
        "token": token,
        "lora_configs": [],
        "add_prompt": False,
        "ddim_steps": 5,
        "init_img": "",
        "n_images": 2,
        "scale": 7,
        "select_seed": -1,
        "enable_lcm": 1,
        "output_size": "1024x1024",
        "cn_configs": [],
        # "model_name": "Artist v0.4.0 Beta",
        "outpaint_config": {
            "src_img": url,
            "outpaint_ratio": 1.5,
            "align_position": "middle"
        }
    }

Artist 3.5 Turbo生图加速示例

{
        "n_images": 1,
        "neg_prompt": "(NSFW:1.5), ((deformed iris, deformed pupils:1.4), ((bad art)),((ugly)),beard, text, image noise, close up, cropped, out of frame, worst quality, low quality, jpeg artifacts, watermark, duplicate, morbid, mutilated, extra fingers, mutated hands, poorly drawn hands, poorly drawn face, mutation, deformed, blurry, dehydrated, bad anatomy, bad proportions, extra limbs, cloned face, disfigured, gross proportions, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck, malformed limbs, missing arms, missing legs, extra arms, extra legs, fused fingers, too many fingers, long neck",
        "select_seed": 1,
        "strength": 0.5,
        "output_size": "640x640",
        "scale": 1.2, //3.5turbo建议1.2
        "add_prompt": "false",
        "token": token,
        "init_img": "",
        "versionid": "sgl_artist_v0.3.5_turbo",//使用该模型进行生图加速
        "ddim_steps": 5,//3.5turbo建议5，步数高生图时间增加
        "prompt": "女孩"
}

结果查询

请求地址

POST https://miaohua.sensetime.com/api/v1b/task_result

请求参数

参数名称 类型 是否必须 默认值 含义
task_id string 是 无 由任务创建接口返回的任务id
token string 是 无 由get_token获取的token
sensitive_limit_value float 否 0 0-1，设置敏感图片鉴定阈值，越大代表越严格，0采用系统默认标准

请求示例

curl 示例

curl https://miaohua.sensetime.com/api/v1b/task_result \
  -H "Content-Type: application/json" \
  -d '{
        "task_id": "f3e5b59c-7416-11ed-a160-00163e025c94",
        "token": "",
  }'

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/task_result'
data = {
    "task_id": "f3e5b59c-7416-11ed-a160-00163e025c94", # string 任务id
    "token": "", # get_token获取的token
}

response = requests.post(url, json=data)

print(response.status_code)
print(response.text)

返回示例

{
    "state": "done", # string 任务状态，未完成会是 "pending"
    'images': [
        {
            'large': "https://pic.bakamaka.io/txt2img-result/2022-12-10/31e17f66-788b-11ed-98e2-00163e025c94_00000.jpg",
            'middle': "https://pic.bakamaka.io/txt2img-result/2022-12-10/31e17f66-788b-11ed-98e2-00163e025c94_00000.mid.jpg",
            'save_index': 0,
            'share': false,
            'shareable': true,
        },
        {
            'large': "https://pic.bakamaka.io/txt2img-result/2022-12-10/31e17f66-788b-11ed-98e2-00163e025c94_00001.jpg",
            'middle': "https://pic.bakamaka.io/txt2img-result/2022-12-10/31e17f66-788b-11ed-98e2-00163e025c94_00001.mid.jpg",
            'save_index': 1,
            'share': false,
            'shareable': true,
        },
        {
            'large': "https://pic.bakamaka.io/txt2img-result/2022-12-10/31e17f66-788b-11ed-98e2-00163e025c94_00002.jpg",
            'middle': "https://pic.bakamaka.io/txt2img-result/2022-12-10/31e17f66-788b-11ed-98e2-00163e025c94_00002.mid.jpg",
            'save_index': 2,
            'share': false,
            'shareable': true,
        },
        {
            'large': "https://pic.bakamaka.io/txt2img-result/2022-12-10/31e17f66-788b-11ed-98e2-00163e025c94_00003.jpg",
            'middle': "https://pic.bakamaka.io/txt2img-result/2022-12-10/31e17f66-788b-11ed-98e2-00163e025c94_00003.mid.jpg",
            'save_index': 3,
            'share': false,
            'shareable': true,
        },
    ], # list 图片路径
    "select_seed": 8206464,#种子数
      
    "error_msg": "Intern error", # string 错误信息
}

批量结果查询

请求地址

POST https://miaohua.sensetime.com/api/v1b/task_results

请求参数

  • 一次可获取多个任务的结果，一次最多不超过50个任务
  • 接口上限为6次/秒，根据实际情况调用
参数名称 类型 是否必须 默认值 含义
task_ids list 是 无 由任务创建接口返回的任务id数组
token string 是 无 由get_token获取的token
sensitive_limit_value float 否 0 0-1，设置敏感图片鉴定阈值，越大代表越严格, 0采用系统默认标准

请求示例

curl 示例

curl https://miaohua.sensetime.com/api/v1b/task_results \
  -H "Content-Type: application/json" \
  -d '{
        "task_ids": [
          "f3e5b59c-7416-11ed-a160-00163e025c94"
        ],
        "token": "",
  }'

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/task_results'
data = {
    "task_ids": ["f3e5b59c-7416-11ed-a160-00163e025c94"], # string 任务id
    "token": "", # get_token获取的token
}

response = requests.post(url, json=data)

print(response.status_code)
print(response.text)

返回示例

{
    "code": 0,
    "info": [
      {
        "state": "done", // string 任务状态，未完成会是 "pending"
        "error_msg": "Intern error", // string 错误信息
        "images": [
            {
                "task_id": "f3e5b59c-7416-11ed-a160-00163e025c94",
                "task_type": "txt2img",
                "large": "https://pic.bakamaka.io/txt2img-result/2022-12-10/31e17f66-788b-11ed-98e2-00163e025c94_00000.jpg",
                "middle": "https://pic.bakamaka.io/txt2img-result/2022-12-10/31e17f66-788b-11ed-98e2-00163e025c94_00000.mid.jpg",
                "small": "https://pic.bakamaka.io/txt2img-result/2022-12-10/31e17f66-788b-11ed-98e2-00163e025c94_00000.mid.jpg",
                "save_index": 0
            }
        ],
        "select_seed": 8206464,#种子数
      },
      {...}
    ]
}

获取生成表单

GET https://miaohua.sensetime.com/api/v1b/get_generation_form

请求参数

无

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/get_generation_form'

response = requests.get(url)

print(response.status_code)
print(response.text)

返回示例

{
  "code": 0,
  "info": [
    {
      "choices": [
        "Artist v0.3.0 Beta",
      ],
      "default": "Artist v0.3.0 Beta",
      "description": "model name",
      "name": "model_name",
      "type": "list"
    },
    {
      "default": "",
      "description": "prompt",
      "name": "prompt",
      "type": "string"
    },
    {
      "default": "",
      "description": "negative prompt",
      "name": "neg_prompt",
      "type": "string"
    },
    {
      "default": 2,
      "description": "number of images",
      "max": 8,
      "min": 1,
      "name": "n_images",
      "type": "int"
    },
    {
      "default": 7,
      "description": "Guidance Scale",
      "max": 20,
      "min": 1,
      "name": "scale",
      "type": "int"
    },
    {
      "default": -1,
      "description": "seed",
      "max": 9999999,
      "min": 0,
      "name": "select_seed",
      "type": "int"
    },
    {
      "choices": {
        "1:1": [
          "512x512",
          "640x640",
          "768x768",
          "1024x1024"
        ],
        "2:3": [
          "512x768",
          "640x960",
          "768x1152",
          "1024x1536"
        ],
        "3:2": [
          "768x512",
          "960x640",
          "1152x768",
          "1536x1024"
        ]
      },
      "default": "512x512",
      "description": "resolution",
      "name": "resolution",
      "type": "dict"
    },
    {
      "default": "",
      "description": "init image url",
      "name": "init_img",
      "type": "str"
    },
    {
      "default": "",
      "description": "controlnet model",
      "name": "controlnet_model",
      "type": "str",
      "choices": ["", "canny"],
    },
  ],
  "msg": ""
}

获取模型：基模型

获取公开的基模型

请求地址

POST https://miaohua.sensetime.com/api/v1b/models/base

请求参数

参数名称 类型 是否必须 默认值 含义
token string 是 无，通过get_token获得 通过get_token获取的token
page int 否 1，第几页 第几页
per_page int 否 20，每页数量 每页数量
query string 否 无，查询 通过模型名、tag等过滤

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/models/base'

response = requests.post(url, json={"token": "xxx"})

print(response.status_code)
print(response.text)

返回示例

{
    "code": 0,
    "info": {
        "has_next": true,
        "models": [
            {
                "author": "xxx",
                "base_model": "xxxx",
                "demo_picture": "https://bkmk.oss-accelerate.aliyuncs.com/8b53e958-374e-11ee-9f88-00163e253f9a_00001_XvYNa9vW_raw.jpeg?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317051652108&Signature=4NyJBRoZk86xleogKqp9ud3M60U%3D",
                "description": "",
                "is_opensource": true,
                "model_id": xx,
                "model_name": "meinamix_meinaV11",
                "tag": "upload",
                "versionid": "xxx"
            }
        ],
        "page": 1,
        "per_page": 1
    },
    "msg": ""
}

获取基模对应的controlnet

请求地址

GET https://miaohua.sensetime.com/api/v1b/cn/list

请求参数

参数名称 类型 是否必须 默认值 含义
versionid string 是 无，通过模型列表接口获得 模型对应versionid

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/cn/list'

response = requests.post(url, json={"token": "token", "versionid":"sgl_artist_v0.3.5"})
print(response.status_code)
print(response.text)
curl -X POST -H "Content-Type: application/json" -d '{"token": "token", "versionid": "sgl_artist_v0.3.5"}' https://miaohua.sensetime.com/api/v1b/cn/list

返回示例

{
   "code":0,
   "info":[
      {
         "can_preview":true,
         "demo_picture":"https://bkmksh.oss-accelerate.aliyuncs.com/87533f22-7950-11ee-b6cd-feb5979cd666.jpeg?OSSAccessKeyId=LTAI5t8GmQec8vxNsiGKcYBT&Expires=317058909727&Signature=UZnEAF%2BMT3kdR1A3Cvi6%2Bigk76U%3D",
         "demo_preview":"None",
         "demo_result":"https://bkmksh.oss-accelerate.aliyuncs.com/8340ac46-464a-11ee-b722-00163e253f9a.jpeg?OSSAccessKeyId=LTAI5t8GmQec8vxNsiGKcYBT&Expires=317058657961&Signature=Jl%2FicV94gpmeiLO3g8A7ShqUqRM%3D",
         "desc":"图像转换：根据用户输入的文本指令，对用户上传的图片进行编辑，转换图片中的部分元素。",
         "local_name":"图像转换",
         "name":"ip2p"
      }
   ],
   "msg":""
}

获取模型：LoRA模型

获取公开的LoRA模型, 可跟基模型lora merge

请求地址

POST https://miaohua.sensetime.com/api/v1b/models/lora

注意： 目前数据正在补齐中，自定义的lora可以同 api/v1b/models/private中获取

请求参数

参数名称 类型 是否必须 默认值 含义
token string 是 无，通过get_token获得 通过get_token获取的token
page int 否 1，第几页 第几页
per_page int 否 20，每页数量 每页数量
query string 否 无，查询 通过模型名、tag等过滤

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/models/lora'

response = requests.post(url, json={"token": "xxx"})

print(response.status_code)
print(response.text)

返回示例

{
    "code": 0,
    "info": {
        "has_next": true,
        "models": [
            {
                "author": "xxx",
                "base_model": "",
                "demo_picture": "https://bkmk.oss-accelerate.aliyuncs.com/8b53e958-374e-11ee-9f88-00163e253f9a_00001_XvYNa9vW_raw.jpeg?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317051652108&Signature=4NyJBRoZk86xleogKqp9ud3M60U%3D",
                "description": "",
                "is_opensource": true,
                "model_id": 5,
                "model_name": "Baby Roy盲盒",
                "tag": "LORA",
                "versionid": "xxx"
            }
        ],
        "page": 1,
        "per_page": 1
    },
    "msg": ""
}

获取模型：私有模型(用户上传/训练)

获取自己上传的基模型/LoRA模型、训练的LoRA模型, 可跟基模型lora merge

请求地址

POST https://miaohua.sensetime.com/api/v1b/models/private

请求参数

参数名称 类型 是否必须 默认值 含义
token string 是 无，通过get_token获得 通过get_token获取的token
page int 否 1，第几页 第几页
per_page int 否 20，每页数量 每页数量
query string 否 无，查询 通过模型名、tag等过滤

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/models/private'

response = requests.post(url, json={"token": "xxx"})

print(response.status_code)
print(response.text)

返回示例

{
    "code": 0,
    "info": {
        "has_next": true,
        "models": [
            {
                "author": "xxx",
                "base_model": "",
                "demo_picture": "https://bkmk.oss-accelerate.aliyuncs.com/8b53e958-374e-11ee-9f88-00163e253f9a_00001_XvYNa9vW_raw.jpeg?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317051652108&Signature=4NyJBRoZk86xleogKqp9ud3M60U%3D",
                "description": "",
                "is_opensource": true,
                "model_id": 5,
                "model_name": "Baby Roy盲盒",
                "tag": "LORA",
                "versionid": "xxx"
            }
        ],
        "page": 1,
        "per_page": 1
    },
    "msg": ""
}

获取Controlnet列表: Artist

获取自研模型支持的controlnet列表， 可用于推理生图控制

请求地址

POST https://miaohua.sensetime.com/api/v1b/cn/artist

请求参数

参数名称 类型 是否必须 默认值 含义
token string 是 无，通过get_token获得 通过get_token获取的token

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/cn/artist'

response = requests.post(url, json={"token": "xxx"})

print(response.status_code)
print(response.text)

返回示例

{
    "code": 0,
    "info": [
        {
            "demo_picture": "https://bkmk.oss-accelerate.aliyuncs.com/75ccd522-3fd7-11ee-b072-00163e253f9a.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317052590512&Signature=UFmUGW1nalt2Ni2HlTkikgV%2Byk4%3D",
            "demo_preview": "https://bkmk.oss-accelerate.aliyuncs.com/b9a6b330-3fd7-11ee-b072-00163e253f9a.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317052590627&Signature=XCcGGQCm%2BmxcDS2pajjBHABjgWo%3D",
            "demo_result": "https://bkmk.oss-accelerate.aliyuncs.com/0e88ca3c-3fd8-11ee-abdf-00163e253f9a.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317052590768&Signature=rvY7Y%2Bd9fgRg5aUvo4WtZC4IJyw%3D",
            "desc": "Onepose: Extracting Character Pose/Pose Recognition can use the figure uploaded by the user as a reference picture for AI painting, extract the pose and movement of the character, and generate a picture of the same pose, with accurate and stunning results.",
            "local_name": "Pose",
            "name": "openpose"
        }
    ],
    "msg": ""
}

获取Controlnet列表: 开源

获取开源模型支持的controlnet列表， 可用于推理生图控制

请求地址

POST https://miaohua.sensetime.com/api/v1b/cn/opensource

请求参数

参数名称 类型 是否必须 默认值 含义
token string 是 无，通过get_token获得 通过get_token获取的token

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/cn/opensource'

response = requests.post(url, json={"token": "xxx"})

print(response.status_code)
print(response.text)

返回示例

{
    "code": 0,
    "info": [
        {
            "demo_picture": "https://bkmk.oss-accelerate.aliyuncs.com/75ccd522-3fd7-11ee-b072-00163e253f9a.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317052590512&Signature=UFmUGW1nalt2Ni2HlTkikgV%2Byk4%3D",
            "demo_preview": "https://bkmk.oss-accelerate.aliyuncs.com/b9a6b330-3fd7-11ee-b072-00163e253f9a.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317052590627&Signature=XCcGGQCm%2BmxcDS2pajjBHABjgWo%3D",
            "demo_result": "https://bkmk.oss-accelerate.aliyuncs.com/0e88ca3c-3fd8-11ee-abdf-00163e253f9a.png?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317052590768&Signature=rvY7Y%2Bd9fgRg5aUvo4WtZC4IJyw%3D",
            "desc": "Onepose: Extracting Character Pose/Pose Recognition can use the figure uploaded by the user as a reference picture for AI painting, extract the pose and movement of the character, and generate a picture of the same pose, with accurate and stunning results.",
            "local_name": "Pose",
            "name": "openpose"
        }
    ],
    "msg": ""
}

以下为训练相关

上传图片

请求地址

POST https://miaohua.sensetime.com/api/v1b/upload_img Authorization: Bearer token_string

请求参数

参数名称 类型 是否必须 默认值 含义
init_img file 是 无 图片文件,为单个，以'init_img'为key

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/upload_imgs'
files={'init_img': open('/local/pic1.png','rb')}
headers = {"Authorization": "Bearer MYREALLYLONGTOKENIGOT"}

response = requests.post(url, files=files, headers=headers)

print(response.status_code)
print(response.json())

返回示例

{
  "code": 0,
  "info": [
    "https://bkmk.oss-accelerate.aliyuncs.com/900ea63e-e1dd-11ed-bef5-00163e005161?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317042257726&Signature=x7nUVN6xpDustx4K02WOTjuty4Q%3D"
  ],
  "msg": ""
}

批量上传图片

请求地址

20230809更新：header中添加Authorization 以识别身份, 在未来版本中将会移除file token

POST https://miaohua.sensetime.com/api/v1b/upload_imgs Authorization: Bearer token_string

请求参数

参数名称 类型 是否必须 默认值 含义
init_img file 是 无 图片文件,可以多个，都以'init_img'为key

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/upload_imgs'
# files的列表格式，以init_img为key
img_paths = ['/local/pic1.png', '/local/pic2.png',]
files = [('init_img', open(img_path, 'rb')) for img_path in img_paths]
headers = {"Authorization": "Bearer MYREALLYLONGTOKENIGOT"}


response = requests.post(url, files=files, headers=headers)

print(response.status_code)
print(response.json())

返回示例

{
  "code": 0,
  "info": [
    "https://bkmk.oss-accelerate.aliyuncs.com/900ea63e-e1dd-11ed-bef5-00163e005161?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317042257726&Signature=x7nUVN6xpDustx4K02WOTjuty4Q%3D",
    "https://bkmk.oss-accelerate.aliyuncs.com/91d5e7d4-e1dd-11ed-bef5-00163e005161?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317042257730&Signature=d1weIkwFwVK9nbCrPNdkeVm1wQA%3D"
  ],
  "msg": ""
}

创建数据集

请求地址

POST https://miaohua.sensetime.com/api/v1b/dataset

请求参数

参数名称 类型 是否必须 默认值 含义
token string 是 无，通过get_token获得 通过get_token获取的token
name string 是 无 数据集名称
description string 否 "" 数据集描述
images list 是 无 图片列表, url list，可通过批量上传图片接口/api/v1b/upload_imgs获取

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/dataset'
data = {
  "token": "xxxx", # get_token获取到的token
  "name": "dataset1", # 数据集名称
  "description": "dataset1 description",  # 数据集描述
  "images": [
    "https://bkmk.oss-accelerate.aliyuncs.com/900ea63e-e1dd-11ed-bef5-00163e005161?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317042257726&Signature=x7nUVN6xpDustx4K02WOTjuty4Q%3D",
    "https://bkmk.oss-accelerate.aliyuncs.com/91d5e7d4-e1dd-11ed-bef5-00163e005161?OSSAccessKeyId=LTAI5tPynodLHeacT1J5SmWh&Expires=317042257730&Signature=d1weIkwFwVK9nbCrPNdkeVm1wQA%3D"
  ] # 图片列表，url list
}

response = requests.post(url, json=data)

print(response.status_code)
print(response.text)

返回示例

{'code': 0, 'info': {'dataset_id': 11}, 'msg': ''}

获取已创建的全部数据集

请求地址

POST https://miaohua.sensetime.com/api/v1b/dataset_all?page=1

url参数

参数名称 类型 是否必须 默认值 含义
page string 否 第几页

请求参数

参数名称 类型 是否必须 默认值 含义
token string 是 无，通过get_token获得 通过get_token获取的token

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/dataset_all'
data = {
  "token": "xxx", # get_token获取到的token
}

response = requests.post(url, json=data)

print(response.status_code)
print(response.text)

返回示例

{
  "items": [
    {
      "id": 1,   # dataset id
      "name": "dataset1",  # dataset 名称
      "description": "dataset1 description",  # dataset 描述
      "demo_picture": "url",   # 封面图
      "create_time": "2023-04-23 00:00:00",  # 创建时间
      "update_time": "2023-04-23 00:00:00",  # 修改时间
      "images": [
        "url1",
        "url2"
      ], # 数据集图片列表
      "images_num": 2 # 数据集图片数量
    },
    {
      "id": 2,
      "name": "dataset2",
      "description": "dataset2 description",
      "demo_picture": "url",
      "create_time": "2023-04-23 00:00:00",
      "update_time": "2023-04-23 00:00:00",
      "images": [
        "url3"
      ],
      "images_num": 1
    }
  ],
  "page": 1,
  "per_page": 20,
  "has_next": false,
  "has_prev": false,
  "totalPages": 1
}

数据集中添加图片

请求地址

POST https://miaohua.sensetime.com/api/v1b/dataset/add_images

请求参数

参数名称 类型 是否必须 默认值 含义
token string 是 无，通过get_token获得 通过get_token获取的token
images array 是 无，图片列表 通过upload_img和upload_imgs得到的oss 图片链接
dataset int 是 无，dataset id 通过/api/v1b/dataset_all获取

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/dataset/add_images'
data = {
  "token": "xxx", # get_token获取到的token
  "dataset": 1,
  "images": [
    "https://bkmk.oss-accelerate.aliyuncs.com/xxxx.JPEG?OSSAccessKeyId=xxx&Signature=xxx"
  ]
}

response = requests.post(url, json=data)

print(response.status_code)
print(response.text)

返回示例

成功

{
    "code": 0,
    "info": "",
    "msg": "OK"
}

失败

{
  "code": 400,
  "info": "",
  "msg": "invalid image"
}

数据集中删除图片

请求地址

POST https://miaohua.sensetime.com/api/v1b/dataset/delete_images

请求参数

参数名称 类型 是否必须 默认值 含义
token string 是 无，通过get_token获得 通过get_token获取的token
images array 是 无，图片列表 通过upload_img和upload_imgs得到的oss 图片链接
dataset int 是 无，dataset id 通过/api/v1b/dataset_all获取

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/dataset/delete_images'
data = {
  "token": "xxx", # get_token获取到的token
  "dataset": 1,
  "images": [
    "https://bkmk.oss-accelerate.aliyuncs.com/xxxx.JPEG?OSSAccessKeyId=xxx&Signature=xxx"
  ]
}

response = requests.post(url, json=data)

print(response.status_code)
print(response.text)

返回示例

成功

{
    "code": 0,
    "info": "",
    "msg": "OK"
}

失败

{
  "code": 400,
  "info": "",
  "msg": "invalid image"
}

提交训练任务

请求地址

POST https://miaohua.sensetime.com/api/v1b/train_submit

请求参数

参数名称 类型 是否必须 默认值 含义
token string 是 无，通过get_token获得 通过get_token获取的token
name string 是 无 训练的LoRA模型名称
description string 否 "" 训练的模型描述
base_model, 废弃字段 string 否 系统指定 🔴❗Warning❗🔴 调用模型训练请改用base_model_version字段， 训练所需要的基本模型，可选模型范围通过/api/v1b/get_train_form获取
base_model_version string 否 系统指定 训练所需要的基本模型versionid，可选模型范围通过/api/v1b/get_train_form获取
trigger_word string 否 "" 触发词
main_body string 是 "Style" 主体， ["Style", "Human", "Architecture", "Plants", "Animals"]
dataset list 是 无 训练所选的数据集id的list,需要先创建数据集

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/train_submit'
data = {
    "token": "", # get_token获取的token
    "name": "train model1", # 训练的模型名称
    "description": "train model1 description",  # 训练的模型描述
    "base_model": "自动选择", # 基模型
    "trigger_word": "one girl", # 触发词
    "main_body": "Style", # 主体
    "dataset": [5, 7], # 选择的数据集的id组合，list
}

response = requests.post(url, json=data)

print(response.status_code)
print(response.text)

返回示例

{'code': 0, 'info': {'task_id': 'xxxxx'}, 'msg': ''}

获取训练进度

请求地址

POST https://miaohua.sensetime.com/api/v1b/train_progress

请求参数

参数名称 类型 是否必须 默认值 含义
token string 是 无，通过get_token获得 通过get_token获取的token
task_id string 是 无 训练任务的task_id

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/train_progress'
data = {
    "token": "", # get_token获取的token
    "task_id": "xxxxxx", # 训练任务的task_id
}

response = requests.post(url, json=data)

print(response.status_code)
print(response.text)

返回示例

{'create_time': '2023-05-08 14:22:52', 'description': 'train model1 description', 'main_body': 'Style', 'model_name': 'train_model1', 'state': 'done', 'training_progress': 1.0, 'trigger_word': 'one girl'}

返回训练state为状态，training_progress为进度(0-1)，状态为'done'后，可通过之前的生成接口生成图片，模型的model_name取训练时取的名字。

获取可供训练的基模列表

请求地址

GET https://miaohua.sensetime.com/api/v1b/get_train_form

请求参数

无

请求示例

python示例

import requests

url = 'https://miaohua.sensetime.com/api/v1b/get_train_form'

response = requests.get(url)

print(response.status_code)
print(response.text)

返回示例

{
  "code": 0,
  "info": {
    "base_model": [
      "商汤自研作画模型 Beta",
      "水彩风",
    ],
    "main_body": [
      "Style",
      "Human",
      "Architecture",
      "Plants",
      "Animals"
    ]
  },
  "msg": ""
}

主要是base_model和main_body这两个字段。