Merge pull request #292 from mrhan1993/support_imagemeta

Support image meta

.gitignore

@@ -1,5 +1,11 @@
+#ide config
+.idea
+.vscode
+
+#runtime
 __pycache__
 .DS_Store
+
 # models
 *.ckpt
 *.safetensors
@@ -22,46 +28,21 @@ conda
 logs
 log
 
-sorted_styles.json
-/language/default.json
-lena.png
-lena_result.png
-lena_test.py
+# config files
+.cog
 config.txt
 config_modification_tutorial.txt
 user_path_config.txt
 user_path_config-deprecated.txt
-build_chb.py
-experiment.py
-/modules/*.png
-/tmp
-/ui-config.json
-/outputs
-/config.json
-/webui.settings.bat
-/embeddings
-/styles.csv
-/params.txt
-/styles.csv.bak
-/webui-user.bat
-/webui-user.sh
-/interrogate
-/user.css
-/.idea
-/notification.ogg
-/notification.mp3
-/SwinIR
-/textual_inversion
-.vscode
-/extensions
-/test/stdout.txt
-/test/stderr.txt
-/cache.json*
-/config_states/
-/node_modules
-/package-lock.json
-/.coverage*
-/auth.json
-.cog/
+
+sorted_styles.json
 /presets
-*.db
+
+# db
+*.db
+
+# cache
+outputs
+
+#other
+*.http

cog.yaml

@@ -37,7 +37,7 @@ build:
     - "colorlog==6.8.2"
     - "rich==13.7.1"
 
-  # commands run after the environment is setup
+  # commands run after the environment is set up
   # run:
   #   - "echo env is ready!"
   #   - "echo another command if needed"

examples/examples_v1.py

@@ -3,7 +3,6 @@
 """
 import json
 import os
-import base64
 import requests
 
 
@@ -12,7 +11,6 @@ class Config:
     Config
     Attributes:
         fooocus_host (str): Fooocus API host
-        text2img_ip (str): Text to Image with IP
         img_upscale (str): Upscale or Vary
         inpaint_outpaint (str): Inpaint or Outpaint
         img_prompt (str): Image Prompt

fooocus_api_version.py

@@ -1 +1 @@
-version = '0.4.0.3'
+version = '0.4.0.4'

fooocusapi/models/common/base.py

@@ -39,7 +39,10 @@ class Lora(BaseModel):
 for lora in default_loras:
     if lora[0] != 'None':
         default_loras_model.append(
-            Lora(enabled=lora[0], model_name=lora[1], weight=lora[2])
+            Lora(
+                enabled=lora[0],
+                model_name=lora[1],
+                weight=lora[2])
         )
 default_loras_json = LoraList.dump_json(default_loras_model)
 
@@ -84,6 +87,16 @@ class DescribeImageType(str, Enum):
     anime = 'Anime'
 
 
+class ImageMetaScheme(str, Enum):
+    """Scheme for save image meta
+    Attributes:
+        Fooocus: json format
+        A111: string
+    """
+    Fooocus = 'fooocus'
+    A111 = 'a111'
+
+
 def style_selection_parser(style_selections: str) -> List[str]:
     """
     Parse style selections, Convert to list
@@ -143,7 +156,8 @@ def outpaint_selections_parser(outpaint_selections: str) -> List[OutpaintExpansi
                     outpaint_selections_arr.append(expansion)
                 except ValueError:
                     errs = InitErrorDetails(
-                        type='enum', loc=['outpaint_selections'],
+                        type='enum',
+                        loc=tuple('outpaint_selections'),
                         input=outpaint_selections,
                         ctx={
                             'expected': "str, comma separated Left, Right, Top, Bottom"
@@ -166,6 +180,9 @@ def image_prompt_parser(image_prompts_config: List[Tuple]) -> List[ImagePrompt]:
         return []
     for config in image_prompts_config:
         cn_img, cn_stop, cn_weight, cn_type = config
-        image_prompts.append(ImagePrompt(cn_img=cn_img, cn_stop=cn_stop,
-                                         cn_weight=cn_weight, cn_type=cn_type))
+        image_prompts.append(ImagePrompt(
+            cn_img=cn_img,
+            cn_stop=cn_stop,
+            cn_weight=cn_weight,
+            cn_type=cn_type))
     return image_prompts

fooocusapi/models/common/image_meta.py

@@ -0,0 +1,116 @@
+"""
+Image meta schema
+"""
+from typing import List
+
+from fooocus_version import version
+from pydantic import BaseModel
+
+
+class ImageMeta(BaseModel):
+    """
+    Image meta data model
+    """
+
+    metadata_scheme: str = "fooocus"
+
+    base_model: str
+    base_model_hash: str
+
+    prompt: str
+    full_prompt: List[str]
+    prompt_expansion: str
+
+    negative_prompt: str
+    full_negative_prompt: List[str]
+
+    performance: str
+
+    style: str
+
+    refiner_model: str = "None"
+    refiner_switch: float = 0.5
+
+    loras: List[list]
+
+    resolution: str
+
+    sampler: str = "dpmpp_2m_sde_gpu"
+    scheduler: str = "karras"
+    seed: str
+    adm_guidance: str
+    guidance_scale: int
+    sharpness: int
+    steps: int
+
+    version: str = version
+
+    def __repr__(self):
+        return ""
+
+
+def loras_parser(loras: list) -> list:
+    """
+    Parse lora list
+    """
+    return [
+        [
+            lora[0].rsplit('.', maxsplit=1)[:1][0],
+            lora[1],
+            "hash_not_calculated",
+        ] for lora in loras if lora[0] != 'None' and lora[0] is not None]
+
+
+def image_parse(
+        async_tak: object,
+        task: dict
+) -> dict | str:
+    """
+    Parse image meta data
+    Generate meta data for image from task and async task object
+    Args:
+        async_tak: async task obj
+        task: task obj
+
+    Returns:
+        dict: image meta data
+    """
+    req_param = async_tak.req_param
+    meta = ImageMeta(
+        metadata_scheme=req_param.meta_scheme,
+        base_model=req_param.base_model_name.rsplit('.', maxsplit=1)[:1][0],
+        base_model_hash='',
+        prompt=req_param.prompt,
+        full_prompt=task['positive'],
+        prompt_expansion=task['expansion'],
+        negative_prompt=req_param.negative_prompt,
+        full_negative_prompt=task['negative'],
+        performance=req_param.performance_selection,
+        style=str(req_param.style_selections),
+        refiner_model=req_param.refiner_model_name,
+        refiner_switch=req_param.refiner_switch,
+        loras=loras_parser(req_param.loras),
+        resolution=str(tuple([int(n) for n in req_param.aspect_ratios_selection.split('*')])),
+        sampler=req_param.advanced_params.sampler_name,
+        scheduler=req_param.advanced_params.scheduler_name,
+        seed=str(task['task_seed']),
+        adm_guidance=str((
+            req_param.advanced_params.adm_scaler_positive,
+            req_param.advanced_params.adm_scaler_negative,
+            req_param.advanced_params.adm_scaler_end)),
+        guidance_scale=req_param.guidance_scale,
+        sharpness=req_param.sharpness,
+        steps=-1,
+        version=version
+    )
+    if meta.metadata_scheme not in ["fooocus", "a111"]:
+        meta.metadata_scheme = "fooocus"
+    if meta.metadata_scheme == "fooocus":
+        meta_dict = meta.model_dump()
+        for i, lora in enumerate(meta.loras):
+            attr_name = f"lora_combined_{i+1}"
+            lr = [str(x) for x in lora]
+            meta_dict[attr_name] = f"{lr[0]} : {lr[1]}"
+    else:
+        meta_dict = meta.model_dump()
+    return meta_dict

fooocusapi/models/common/requests.py

@@ -26,6 +26,7 @@
 
 default_aspect_ratio = default_aspect_ratio.split(" ")[0].replace("×", "*")
 
+
 class QueryJobRequest(BaseModel):
     """Query job request"""
     job_id: str = Field(description="Job ID to query")
@@ -90,7 +91,10 @@ class CommonRequest(BaseModel):
     refiner_switch: float = Field(default=default_refiner_switch, description="Refiner Switch At", ge=0.1, le=1.0)
     loras: List[Lora] = Field(default=default_loras_model)
     advanced_params: AdvancedParams = AdvancedParams()
+    save_meta: bool = Field(default=True, description="Save meta data")
+    meta_scheme: str = Field(defaut='fooocus', description="Meta data scheme, one of [fooocus, a111]")
     save_extension: str = Field(default='png', description="Save extension, one of [png, jpg, webp]")
+    save_name: str = Field(default='', description="Image name for output image, default is a uuid")
     read_wildcards_in_order: bool = Field(default=False, description="Read wildcards in order")
     require_base64: bool = Field(default=False, description="Return base64 data of generated image")
     async_process: bool = Field(default=False, description="Set to true will run async and return job info for retrieve generation result later")

fooocusapi/parameters.py

@@ -165,6 +165,9 @@ def __init__(
         image_prompts: List[Tuple[np.ndarray, float, float, str]],
         advanced_params: List[any] | None,
         save_extension: str,
+        save_meta: bool,
+        meta_scheme: str,
+        save_name: str,
         require_base64: bool,
     ):
         self.prompt = prompt
@@ -192,6 +195,9 @@ def __init__(
         self.inpaint_additional_prompt = inpaint_additional_prompt
         self.image_prompts = image_prompts
         self.save_extension = save_extension
+        self.save_meta = save_meta
+        self.meta_scheme = meta_scheme
+        self.save_name = save_name
         self.require_base64 = require_base64
         self.advanced_params = advanced_params
 

fooocusapi/utils/api_utils.py

@@ -204,6 +204,9 @@ def req_to_params(req: Text2ImgRequest) -> ImageGenerationParams:
         inpaint_additional_prompt=inpaint_additional_prompt,
         image_prompts=image_prompts,
         advanced_params=advanced_params,
+        save_meta=req.save_meta,
+        meta_scheme=req.meta_scheme,
+        save_name=req.save_name,
         save_extension=req.save_extension,
         require_base64=req.require_base64,
     )

fooocusapi/utils/file_utils.py

@@ -29,8 +29,11 @@
 STATIC_SERVER_BASE = 'http://127.0.0.1:8888/files/'
 
 
-def save_output_file(img: np.ndarray, image_meta: dict = None,
-                     image_name: str = '', extension: str = 'png') -> str:
+def save_output_file(
+        img: np.ndarray,
+        image_meta: dict = None,
+        image_name: str = '',
+        extension: str = 'png') -> str:
     """
     Save np image to file
     Args:
@@ -44,8 +47,6 @@ def save_output_file(img: np.ndarray, image_meta: dict = None,
     current_time = datetime.datetime.now()
     date_string = current_time.strftime("%Y-%m-%d")
 
-    image_name = str(uuid.uuid4()) if image_name == '' else image_name
-
     filename = os.path.join(date_string, image_name + '.' + extension)
     file_path = os.path.join(output_dir, filename)
 
@@ -58,7 +59,8 @@ def save_output_file(img: np.ndarray, image_meta: dict = None,
     meta = None
     if extension == 'png':
         meta = PngInfo()
-        meta.add_text("params", json.dumps(image_meta))
+        meta.add_text("parameters", json.dumps(image_meta))
+        meta.add_text("fooocus_scheme", image_meta['metadata_scheme'])
 
     os.makedirs(os.path.dirname(file_path), exist_ok=True)
     Image.fromarray(img).save(file_path, format=extension,