Face2Doll (U2Net)

.gitattributes

@@ -25,3 +25,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zstandard filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+u2net_bce_itr_18000_train_3.891670_tar_0.553700_512x_460x.jit.pt filter=lfs diff=lfs merge=lfs -text
+shape_predictor_5_face_landmarks.dat filter=lfs diff=lfs merge=lfs -text

README.md

@@ -1,8 +1,8 @@
 ---
-title: Face2Doll
-emoji: 🌖
-colorFrom: purple
-colorTo: green
+title: Face2Doll (U2Net)
+emoji: 🎎
+colorFrom: pink
+colorTo: black
 sdk: gradio
 app_file: app.py
 pinned: false

app.py

@@ -0,0 +1,112 @@
+import os
+os.system("pip install dlib")
+os.system("pip install gradio==2.5.3")
+import sys
+import face_detection
+import PIL
+from PIL import Image, ImageOps, ImageFile
+import numpy as np
+
+import torch
+torch.set_grad_enabled(False)
+model = torch.jit.load('u2net_bce_itr_18000_train_3.891670_tar_0.553700_512x_460x.jit.pt')
+model.eval()
+
+def normPRED(d):
+    ma = np.max(d)
+    mi = np.min(d)
+
+    dn = (d-mi)/(ma-mi)
+
+    return dn
+
+def array_to_image(array_in):    
+    array_in = normPRED(array_in)
+    array_in = np.squeeze(255.0*(array_in)) 
+    array_in = np.transpose(array_in, (1, 2, 0))
+    im = Image.fromarray(array_in.astype(np.uint8))
+    return im
+
+
+def image_as_array(image_in):
+    image_in = np.array(image_in, np.float32)
+    tmpImg = np.zeros((image_in.shape[0],image_in.shape[1],3))
+    image_in = image_in/np.max(image_in)
+    if image_in.shape[2]==1:
+        tmpImg[:,:,0] = (image_in[:,:,0]-0.485)/0.229
+        tmpImg[:,:,1] = (image_in[:,:,0]-0.485)/0.229
+        tmpImg[:,:,2] = (image_in[:,:,0]-0.485)/0.229
+    else:
+        tmpImg[:,:,0] = (image_in[:,:,0]-0.485)/0.229
+        tmpImg[:,:,1] = (image_in[:,:,1]-0.456)/0.224
+        tmpImg[:,:,2] = (image_in[:,:,2]-0.406)/0.225
+
+    tmpImg = tmpImg.transpose((2, 0, 1))
+    image_out = np.expand_dims(tmpImg, 0)
+    return image_out
+
+def find_aligned_face(image_in, size=512):   
+    aligned_image, n_faces, quad = face_detection.align(image_in, face_index=0, output_size=size)
+    return aligned_image, n_faces, quad
+
+def align_first_face(image_in, size=512):
+    aligned_image, n_faces, quad = find_aligned_face(image_in,size=size)
+    if n_faces == 0:
+        try:
+            image_in = ImageOps.exif_transpose(image_in)
+        except:
+            print("exif problem, not rotating")
+        image_in = image_in.resize((size, size))
+        im_array = image_as_array(image_in)
+    else:
+        im_array = image_as_array(aligned_image)
+
+    return im_array
+
+def img_concat_h(im1, im2):
+    dst = Image.new('RGB', (im1.width + im2.width, im1.height))
+    dst.paste(im1, (0, 0))
+    dst.paste(im2, (im1.width, 0))
+    return dst
+
+import gradio as gr
+
+def face2doll(
+    img: Image.Image,
+    size: int
+) -> Image.Image:
+
+    aligned_img = align_first_face(img)
+    if aligned_img is None:
+        output=None
+    else:
+        input = torch.Tensor(aligned_img)
+        results = model(input)
+        d2 = array_to_image(results[1].detach().numpy())
+        output = img_concat_h(array_to_image(aligned_img), d2)         
+        del results
+
+    return output
+
+def inference(img):
+    out = face2doll(img, 512)
+    return out
+      
+  
+title = "Face2Doll U2Net"
+description = "Style transfer a face into one of a \"Doll\". Upload an image with a face, or click on one of the examples below. If a face could not be detected, an image will still be created. Faces with glasses on, seem not to yield good results."
+article = "<hr><p style='text-align: center'>See the <a href='https://github.com/Norod/U-2-Net-StyleTransfer' target='_blank'>Github Repo</a></p><p style='text-align: center'>samples: <img src='https://hf.space/gradioiframe/Norod78/Face2Doll/file/Sample00001.jpg' alt='Sample00001'/><img src='https://hf.space/gradioiframe/Norod78/Face2Doll/file/Sample00002.jpg' alt='Sample00002'/><img src='https://hf.space/gradioiframe/Norod78/Face2Doll/file/Sample00003.jpg' alt='Sample00003'/><img src='https://hf.space/gradioiframe/Norod78/Face2Doll/file/Sample00004.jpg' alt='Sample00004'/><img src='https://hf.space/gradioiframe/Norod78/Face2Doll/file/Sample00005.jpg' alt='Sample00005'/></p><p>The \"Face2Doll (U2Net)\" model was trained by <a href='https://linktr.ee/Norod78' target='_blank'>Doron Adler</a></p>"
+
+examples=[['Example00001.jpg'],['Example00002.jpg'],['Example00003.jpg'],['Example00004.jpg'],['Example00005.jpg'], ['Example00006.jpg']]
+
+gr.Interface(
+    inference, 
+    gr.inputs.Image(type="pil", label="Input"), 
+    gr.outputs.Image(type="pil", label="Output"),
+    title=title,
+    description=description,
+    article=article,
+    examples=examples,
+    enable_queue=True,
+    allow_flagging=False
+    ).launch()

face_detection.py

@@ -0,0 +1,140 @@
+# Copyright (c) 2021 Justin Pinkney
+
+import dlib
+import numpy as np
+import os
+from PIL import Image
+from PIL import ImageOps
+from scipy.ndimage import gaussian_filter
+import cv2
+
+
+MODEL_PATH = "shape_predictor_5_face_landmarks.dat"
+detector = dlib.get_frontal_face_detector()
+
+
+def align(image_in, face_index=0, output_size=256):
+    try:
+        image_in = ImageOps.exif_transpose(image_in)
+    except:
+        print("exif problem, not rotating")
+
+    landmarks = list(get_landmarks(image_in))
+    n_faces = len(landmarks)
+    face_index = min(n_faces-1, face_index)
+    if n_faces == 0:
+        aligned_image = image_in
+        quad = None
+    else:
+        aligned_image, quad = image_align(image_in, landmarks[face_index], output_size=output_size)
+
+    return aligned_image, n_faces, quad
+
+
+def composite_images(quad, img, output):
+    """Composite an image into and output canvas according to transformed co-ords"""
+    output = output.convert("RGBA")
+    img = img.convert("RGBA")
+    input_size = img.size
+    src = np.array(((0, 0), (0, input_size[1]), input_size, (input_size[0], 0)), dtype=np.float32)
+    dst = np.float32(quad)
+    mtx = cv2.getPerspectiveTransform(dst, src)
+    img = img.transform(output.size, Image.PERSPECTIVE, mtx.flatten(), Image.BILINEAR)
+    output.alpha_composite(img)
+
+    return output.convert("RGB")
+    
+
+def get_landmarks(image):
+    """Get landmarks from PIL image"""
+    shape_predictor = dlib.shape_predictor(MODEL_PATH)
+
+    max_size = max(image.size)
+    reduction_scale = int(max_size/512)
+    if reduction_scale == 0:
+        reduction_scale = 1
+    downscaled = image.reduce(reduction_scale)
+    img = np.array(downscaled)
+    detections = detector(img, 0)
+    
+    for detection in detections:
+        try:
+            face_landmarks = [(reduction_scale*item.x, reduction_scale*item.y) for item in shape_predictor(img, detection).parts()]
+            yield face_landmarks
+        except Exception as e:
+            print(e)
+
+
+def image_align(src_img, face_landmarks, output_size=512, transform_size=2048, enable_padding=True, x_scale=1, y_scale=1, em_scale=0.1, alpha=False):
+        # Align function modified from ffhq-dataset
+        # See https://github.com/NVlabs/ffhq-dataset for license
+
+        lm = np.array(face_landmarks)
+        lm_eye_left      = lm[2:3]  # left-clockwise
+        lm_eye_right     = lm[0:1]  # left-clockwise
+
+        # Calculate auxiliary vectors.
+        eye_left     = np.mean(lm_eye_left, axis=0)
+        eye_right    = np.mean(lm_eye_right, axis=0)
+        eye_avg      = (eye_left + eye_right) * 0.5
+        eye_to_eye   = 0.71*(eye_right - eye_left)
+        mouth_avg    = lm[4]
+        eye_to_mouth = 1.35*(mouth_avg - eye_avg)
+
+        # Choose oriented crop rectangle.
+        x = eye_to_eye.copy()
+        x /= np.hypot(*x)
+        x *= max(np.hypot(*eye_to_eye) * 2.0, np.hypot(*eye_to_mouth) * 1.8)
+        x *= x_scale
+        y = np.flipud(x) * [-y_scale, y_scale]
+        c = eye_avg + eye_to_mouth * em_scale
+        quad = np.stack([c - x - y, c - x + y, c + x + y, c + x - y])
+        quad_orig = quad.copy()
+        qsize = np.hypot(*x) * 2        
+        
+        img = src_img.convert('RGBA').convert('RGB')
+
+        # Shrink.
+        shrink = int(np.floor(qsize / output_size * 0.5))
+        if shrink > 1:
+            rsize = (int(np.rint(float(img.size[0]) / shrink)), int(np.rint(float(img.size[1]) / shrink)))
+            img = img.resize(rsize, Image.ANTIALIAS)
+            quad /= shrink
+            qsize /= shrink
+
+        # Crop.
+        border = max(int(np.rint(qsize * 0.1)), 3)
+        crop = (int(np.floor(min(quad[:,0]))), int(np.floor(min(quad[:,1]))), int(np.ceil(max(quad[:,0]))), int(np.ceil(max(quad[:,1]))))
+        crop = (max(crop[0] - border, 0), max(crop[1] - border, 0), min(crop[2] + border, img.size[0]), min(crop[3] + border, img.size[1]))
+        if crop[2] - crop[0] < img.size[0] or crop[3] - crop[1] < img.size[1]:
+            img = img.crop(crop)
+            quad -= crop[0:2]
+
+        # Pad.
+        pad = (int(np.floor(min(quad[:,0]))), int(np.floor(min(quad[:,1]))), int(np.ceil(max(quad[:,0]))), int(np.ceil(max(quad[:,1]))))
+        pad = (max(-pad[0] + border, 0), max(-pad[1] + border, 0), max(pad[2] - img.size[0] + border, 0), max(pad[3] - img.size[1] + border, 0))
+        if enable_padding and max(pad) > border - 4:
+            pad = np.maximum(pad, int(np.rint(qsize * 0.3)))
+            img = np.pad(np.float32(img), ((pad[1], pad[3]), (pad[0], pad[2]), (0, 0)), 'reflect')
+            h, w, _ = img.shape
+            y, x, _ = np.ogrid[:h, :w, :1]
+            mask = np.maximum(1.0 - np.minimum(np.float32(x) / pad[0], np.float32(w-1-x) / pad[2]), 1.0 - np.minimum(np.float32(y) / pad[1], np.float32(h-1-y) / pad[3]))
+            blur = qsize * 0.02
+            img += (gaussian_filter(img, [blur, blur, 0]) - img) * np.clip(mask * 3.0 + 1.0, 0.0, 1.0)
+            img += (np.median(img, axis=(0,1)) - img) * np.clip(mask, 0.0, 1.0)
+            img = np.uint8(np.clip(np.rint(img), 0, 255))
+            if alpha:
+                mask = 1-np.clip(3.0 * mask, 0.0, 1.0)
+                mask = np.uint8(np.clip(np.rint(mask*255), 0, 255))
+                img = np.concatenate((img, mask), axis=2)
+                img = Image.fromarray(img, 'RGBA')
+            else:
+                img = Image.fromarray(img, 'RGB')
+            quad += pad[:2]
+
+        # Transform.
+        img = img.transform((transform_size, transform_size), Image.QUAD, (quad + 0.5).flatten(), Image.BILINEAR)
+        if output_size < transform_size:
+            img = img.resize((output_size, output_size), Image.ANTIALIAS)
+
+        return img, quad_orig

requirements.txt

@@ -0,0 +1,8 @@
+numpy
+opencv-python-headless
+Pillow
+scikit-image
+torch
+torchvision
+scipy
+cmake

shape_predictor_5_face_landmarks.dat

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c4b1e9804792707d3a405c2c16a80a20269e6675021f64a41d30fffafbc41888
+size 9150489

u2net_bce_itr_18000_train_3.891670_tar_0.553700_512x_460x.jit.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4759db0de410078514769bdae185cf80b095e8a5b49aef31987ad5fa6e66a30
+size 177290264