feat(backend): fix nsfw checker catch-22

invokeai/app/api/routers/app_info.py

@@ -13,7 +13,6 @@
 from invokeai.app.invocations.upscale import ESRGAN_MODELS
 from invokeai.app.services.invocation_cache.invocation_cache_common import InvocationCacheStatus
 from invokeai.backend.image_util.infill_methods.patchmatch import PatchMatch
-from invokeai.backend.image_util.safety_checker import SafetyChecker
 from invokeai.backend.util.logging import logging
 from invokeai.version import __version__
 
@@ -109,9 +108,7 @@ async def get_config() -> AppConfig:
         upscaling_models.append(str(Path(model).stem))
     upscaler = Upscaler(upscaling_method="esrgan", upscaling_models=upscaling_models)
 
-    nsfw_methods = []
-    if SafetyChecker.safety_checker_available():
-        nsfw_methods.append("nsfw_checker")
+    nsfw_methods = ["nsfw_checker"]
 
     watermarking_methods = ["invisible_watermark"]
 

invokeai/app/invocations/image.py

@@ -1,6 +1,5 @@
 # Copyright (c) 2022 Kyle Schouviller (https://github.com/kyle0654)
 
-from pathlib import Path
 from typing import Literal, Optional
 
 import cv2
@@ -504,7 +503,7 @@ def invoke(self, context: InvocationContext) -> ImageOutput:
     title="Blur NSFW Image",
     tags=["image", "nsfw"],
     category="image",
-    version="1.2.2",
+    version="1.2.3",
 )
 class ImageNSFWBlurInvocation(BaseInvocation, WithMetadata, WithBoard):
     """Add blur to NSFW-flagged images"""
@@ -516,23 +515,12 @@ def invoke(self, context: InvocationContext) -> ImageOutput:
 
         logger = context.logger
         logger.debug("Running NSFW checker")
-        if SafetyChecker.has_nsfw_concept(image):
-            logger.info("A potentially NSFW image has been detected. Image will be blurred.")
-            blurry_image = image.filter(filter=ImageFilter.GaussianBlur(radius=32))
-            caution = self._get_caution_img()
-            blurry_image.paste(caution, (0, 0), caution)
-            image = blurry_image
+        image = SafetyChecker.blur_if_nsfw(image)
 
         image_dto = context.images.save(image=image)
 
         return ImageOutput.build(image_dto)
 
-    def _get_caution_img(self) -> Image.Image:
-        import invokeai.app.assets.images as image_assets
-
-        caution = Image.open(Path(image_assets.__path__[0]) / "caution.png")
-        return caution.resize((caution.width // 2, caution.height // 2))
-
 
 @invocation(
     "img_watermark",

invokeai/backend/image_util/safety_checker.py

@@ -8,14 +8,15 @@
 
 import numpy as np
 from diffusers.pipelines.stable_diffusion.safety_checker import StableDiffusionSafetyChecker
-from PIL import Image
+from PIL import Image, ImageFilter
 from transformers import AutoFeatureExtractor
 
 import invokeai.backend.util.logging as logger
 from invokeai.app.services.config.config_default import get_config
 from invokeai.backend.util.devices import TorchDevice
 from invokeai.backend.util.silence_warnings import SilenceWarnings
 
+repo_id = "CompVis/stable-diffusion-safety-checker"
 CHECKER_PATH = "core/convert/stable-diffusion-safety-checker"
 
 
@@ -24,30 +25,30 @@ class SafetyChecker:
     Wrapper around SafetyChecker model.
     """
 
-    safety_checker = None
     feature_extractor = None
-    tried_load: bool = False
+    safety_checker = None
 
     @classmethod
     def _load_safety_checker(cls):
-        if cls.tried_load:
+        if cls.safety_checker is not None and cls.feature_extractor is not None:
             return
 
         try:
-            cls.safety_checker = StableDiffusionSafetyChecker.from_pretrained(get_config().models_path / CHECKER_PATH)
-            cls.feature_extractor = AutoFeatureExtractor.from_pretrained(get_config().models_path / CHECKER_PATH)
+            model_path = get_config().models_path / CHECKER_PATH
+            if model_path.exists():
+                cls.feature_extractor = AutoFeatureExtractor.from_pretrained(model_path)
+                cls.safety_checker = StableDiffusionSafetyChecker.from_pretrained(model_path)
+            else:
+                model_path.mkdir(parents=True, exist_ok=True)
+                cls.feature_extractor = AutoFeatureExtractor.from_pretrained(repo_id)
+                cls.feature_extractor.save_pretrained(model_path, safe_serialization=True)
+                cls.safety_checker = StableDiffusionSafetyChecker.from_pretrained(repo_id)
+                cls.safety_checker.save_pretrained(model_path, safe_serialization=True)
         except Exception as e:
             logger.warning(f"Could not load NSFW checker: {str(e)}")
-        cls.tried_load = True
-
-    @classmethod
-    def safety_checker_available(cls) -> bool:
-        return Path(get_config().models_path, CHECKER_PATH).exists()
 
     @classmethod
     def has_nsfw_concept(cls, image: Image.Image) -> bool:
-        if not cls.safety_checker_available() and cls.tried_load:
-            return False
         cls._load_safety_checker()
         if cls.safety_checker is None or cls.feature_extractor is None:
             return False
@@ -60,3 +61,24 @@ def has_nsfw_concept(cls, image: Image.Image) -> bool:
         with SilenceWarnings():
             checked_image, has_nsfw_concept = cls.safety_checker(images=x_image, clip_input=features.pixel_values)
         return has_nsfw_concept[0]
+
+    @classmethod
+    def blur_if_nsfw(cls, image: Image.Image) -> Image.Image:
+        if cls.has_nsfw_concept(image):
+            logger.warning("A potentially NSFW image has been detected. Image will be blurred.")
+            blurry_image = image.filter(filter=ImageFilter.GaussianBlur(radius=32))
+            caution = cls._get_caution_img()
+            # Center the caution image on the blurred image
+            x = (blurry_image.width - caution.width) // 2
+            y = (blurry_image.height - caution.height) // 2
+            blurry_image.paste(caution, (x, y), caution)
+            image = blurry_image
+
+        return image
+
+    @classmethod
+    def _get_caution_img(cls) -> Image.Image:
+        import invokeai.app.assets.images as image_assets
+
+        caution = Image.open(Path(image_assets.__path__[0]) / "caution.png")
+        return caution.resize((caution.width // 2, caution.height // 2))