Merge pull request #150 from ototadana/add-example-mask-images

add example mask images
ototadana · Aug 13, 2023 · fffa3b7 · fffa3b7
2 parents 0ecfa42 + 7f339e7
commit fffa3b7
Show file tree

Hide file tree

Showing 19 changed files with 31 additions and 0 deletions.
diff --git a/images/inferencers/anime_segmentation/mask.jpg b/images/inferencers/anime_segmentation/mask.jpg
diff --git a/images/inferencers/bisenet-face-hair-hat-neck.jpg b/images/inferencers/bisenet-face-hair-hat-neck.jpg
diff --git a/images/inferencers/bisenet-face-hair-hat.jpg b/images/inferencers/bisenet-face-hair-hat.jpg
diff --git a/images/inferencers/bisenet-face-hair.jpg b/images/inferencers/bisenet-face-hair.jpg
diff --git a/images/inferencers/bisenet-face.jpg b/images/inferencers/bisenet-face.jpg
diff --git a/images/inferencers/blur.jpg b/images/inferencers/blur.jpg
diff --git a/images/inferencers/ellipse.jpg b/images/inferencers/ellipse.jpg
diff --git a/images/inferencers/face-area-rect.jpg b/images/inferencers/face-area-rect.jpg
diff --git a/images/inferencers/insightface/mask.jpg b/images/inferencers/insightface/mask.jpg
diff --git a/images/inferencers/mediapipe/mask.jpg b/images/inferencers/mediapipe/mask.jpg
diff --git a/images/inferencers/no-mask.jpg b/images/inferencers/no-mask.jpg
diff --git a/images/inferencers/openmmlab/mask.jpg b/images/inferencers/openmmlab/mask.jpg
diff --git a/images/inferencers/vignette-default.jpg b/images/inferencers/vignette-default.jpg
diff --git a/images/inferencers/vignette-use-minimal-area.jpg b/images/inferencers/vignette-use-minimal-area.jpg
diff --git a/scripts/inferencers/README.md b/scripts/inferencers/README.md
@@ -83,6 +83,8 @@ This is the implementation used for enhancing enlarged face images in the defaul
 ### 2.2 Blur
 This face processor applies a Gaussian blur to the detected face region. The intensity of the blur can be specified using the `radius` parameter in the 'params' of the JSON configuration. The larger the radius, the more intense the blur effect.
 
+![Blur](../../images/inferencers/blur.jpg)
+
 #### Name
 - Blur
 
@@ -127,6 +129,19 @@ This operates as the Mask Generator in the default workflow. Similar to RetinaFa
 
 This component is implemented using [facexlib/parsing](https://github.com/xinntao/facexlib/blob/master/facexlib/parsing/__init__.py).
 
+#### Affected areas - Face
+![Face](../../images/inferencers/bisenet-face.jpg)
+
+#### Affected areas - Face, Hair
+![Face, Hair](../../images/inferencers/bisenet-face-hair.jpg)
+
+#### Affected areas - Face, Hair, Hat
+![Face, Hair, Hat](../../images/inferencers/bisenet-face-hair-hat.jpg)
+
+#### Affected areas - Face, Hair, Hat, Neck
+![Face, Hair, Hat, Neck](../../images/inferencers/bisenet-face-hair-hat-neck.jpg)
+
+
 #### Name
 - BiSeNet
 
@@ -149,6 +164,8 @@ This component is implemented using [facexlib/parsing](https://github.com/xinnta
 ### 3.2 Vignette
 This mask generator creates a mask by applying a Gaussian (circular fade-out effect) to the face area. It is less computationally demanding than deep-learning-based mask generators and can consistently produce a mask under conditions where deep-learning-based mask generators such as BiSeNet or YOLO may struggle, such as with unusual face orientations or expressions. It serves as the fallback mask generator for the BiSeNet Mask Generator when it fails to generate an appropriate mask. 
 
+![Vignette](../../images/inferencers/vignette-default.jpg)
+
 #### Name
 - Vignette
 
@@ -167,6 +184,8 @@ This mask generator creates a mask by applying a Gaussian (circular fade-out eff
 ### 3.3 Ellipse
 This option draws an ellipse around the detected face region to generate a mask.
 
+![Ellipse](../../images/inferencers/ellipse.jpg)
+
 #### Name
 - Ellipse
 
@@ -184,6 +203,8 @@ This option draws an ellipse around the detected face region to generate a mask.
 ### 3.4 Rect
 This is a simplistic implementation that uses the detected face region as a direct mask.
 
+![Rect](../../images/inferencers/face-area-rect.jpg)
+
 #### Name
 - Rect
 
@@ -201,6 +222,8 @@ This is a simplistic implementation that uses the detected face region as a dire
 ### 3.5 NoMask
 This option generates a "mask" that is simply an all-white image of the same size as the input face image. It essentially does not mask any part of the image and can be used in scenarios where no masking is desired.
 
+![NoMask](../../images/inferencers/no-mask.jpg)
+
 #### Implementation
 - [NoMaskGenerator](no_mask_generator.py)
 

diff --git a/scripts/inferencers/anime_segmentation/README.md b/scripts/inferencers/anime_segmentation/README.md
@@ -6,6 +6,8 @@ To use the following components, please enable 'anime_segmentation' option under
 ## 1. Mask Generator
 This utilizes the [Anime Segmentation](https://github.com/SkyTNT/anime-segmentation) model to generate masks specifically designed for anime images.
 
+![Example](../../../images//inferencers/anime_segmentation/mask.jpg)
+
 #### Name
 - AnimeSegmentation
 

diff --git a/scripts/inferencers/insightface/README.md b/scripts/inferencers/insightface/README.md
@@ -38,6 +38,8 @@ Please be aware that the accuracy of gender and age detection with the InsightFa
 ## 2. Mask Generator
 This component utilizes the facial landmark detection feature of InsightFace to generate masks. It identifies the 106 facial landmarks provided by the InsightFace's 'landmark_2d_106' model and uses them to construct the mask.
 
+![Example](../../../images/inferencers/insightface/mask.jpg)
+
 #### Name
 - InsightFace
 

diff --git a/scripts/inferencers/mediapipe/README.md b/scripts/inferencers/mediapipe/README.md
@@ -32,6 +32,8 @@ A Face Detector implemented using [the face detection feature of MediaPipe](http
 ## 2. Mask Generator
 This mask generator uses the MediaPipe Face Mesh model to generate masks. It identifies a set of facial landmarks and interpolates these to create a mask.
 
+![Example](../../../images/inferencers/mediapipe/mask.jpg)
+
 #### Name
 - MediaPipe
 

diff --git a/scripts/inferencers/openmmlab/README.md b/scripts/inferencers/openmmlab/README.md
@@ -6,6 +6,8 @@ To use this, please enable 'openmmlab' option under "Additional components" in t
 ## 1. Mask Generator
 This component utilizes a [Occlusion-aware face segmentation model](https://huggingface.co/ototadana/occlusion-aware-face-segmentation) for mask generation. Because the model is trained by [High-Quality Synthetic Face Occlusion Segmentation Datasets](https://github.com/kennyvoo/face-occlusion-generation), if a face is covered by another object, such as a hand or mask, it can be excluded and the facial segment can be extracted.
 
+![Example](../../../images/inferencers/openmmlab/mask.jpg)
+
 #### Name
 - MMSeg