update

.gitattributes

@@ -4,3 +4,4 @@ hivision/creator/weights/hivision_modnet.onnx filter=lfs diff=lfs merge=lfs -tex
 hivision/creator/weights/modnet_photographic_portrait_matting.onnx filter=lfs diff=lfs merge=lfs -text
 hivision/creator/weights/rmbg-1.4.onnx filter=lfs diff=lfs merge=lfs -text
 hivision/creator/retinaface/weights/retinaface-resnet50.onnx filter=lfs diff=lfs merge=lfs -text
+demo/assets/american-style.png filter=lfs diff=lfs merge=lfs -text

demo/locales.py

@@ -182,22 +182,22 @@ LOCALES = {
     "bg_color": {
         "en": {
             "label": "Background color",
-            "choices": list(color_list_dict_EN.keys()) + ["Custom(RGB)", "Custom(HEX)"],
+            "choices": list(color_list_dict_EN.keys()) + ["American Style"] + ["Custom(RGB)", "Custom(HEX)"],
             "develop": color_list_dict_EN,
         },
         "zh": {
             "label": "背景颜色",
-            "choices": list(color_list_dict_CN.keys()) + ["自定义(RGB)", "自定义(HEX)"],
+            "choices": list(color_list_dict_CN.keys()) + ["美式证件照"] + ["自定义(RGB)", "自定义(HEX)"],
             "develop": color_list_dict_CN,
         },
         "ja": {
             "label": "背景色",
-            "choices": list(color_list_dict_EN.keys()) + ["カスタム(RGB)", "カスタム(HEX)"],
+            "choices": list(color_list_dict_EN.keys()) + ["American Style"] + ["カスタム(RGB)", "カスタム(HEX)"],
             "develop": color_list_dict_EN,
         },
         "ko": {
             "label": "배경색",
-            "choices": list(color_list_dict_EN.keys()) + ["사용자 지정(RGB)", "사용자 지정(HEX)"],
+            "choices": list(color_list_dict_EN.keys()) + ["American Style"] + ["사용자 지정(RGB)", "사용자 지정(HEX)"],
             "develop": color_list_dict_EN,
         },
     },

demo/processor.py

@@ -3,6 +3,7 @@ from hivision import IDCreator
 from hivision.error import FaceError, APIError
 from hivision.utils import (
     add_background,
+    add_background_with_image,
     resize_image_to_kb,
     add_watermark,
     save_image_dpi_to_bytes,
@@ -16,10 +17,13 @@ from hivision.plugin.template.template_calculator import generte_template_photo
 from demo.utils import range_check
 import gradio as gr
 import os
+import cv2
 import time
 from demo.locales import LOCALES
 
 
+base_path = os.path.dirname(os.path.abspath(__file__))
+
 class IDPhotoProcessor:
     def process(
         self,
@@ -217,11 +221,12 @@ class IDPhotoProcessor:
         custom_color_hex_value,
     ):
         """处理颜色模式"""
-        # 如果选择了自定义颜色
+        # 如果选择了自定义颜色BGR
         if idphoto_json["color_mode"] == LOCALES["bg_color"][language]["choices"][-2]:
             idphoto_json["color_bgr"] = tuple(
                 map(range_check, [custom_color_R, custom_color_G, custom_color_B])
             )
+        # 如果选择了自定义颜色HEX
         elif idphoto_json["color_mode"] == LOCALES["bg_color"][language]["choices"][-1]:
             hex_color = custom_color_hex_value
             # 将十六进制颜色转换为RGB颜色，如果长度为6，则直接转换，如果长度为7，则去掉#号再转换
@@ -238,6 +243,9 @@ class IDPhotoProcessor:
                 raise ValueError(
                     "Invalid hex color. You can only use 6 or 7 characters. For example: #FFFFFF or FFFFFF"
                 )
+        # 如果选择了美式证件照
+        elif idphoto_json["color_mode"] == LOCALES["bg_color"][language]["choices"][-3]:
+            idphoto_json["color_bgr"] = (255, 255, 255)
         else:
             hex_color = LOCALES["bg_color"][language]["develop"][color_option]
             idphoto_json["color_bgr"] = tuple(
@@ -311,7 +319,7 @@ class IDPhotoProcessor:
 
         # 渲染背景
         result_image_standard, result_image_hd = self._render_background(
-            result_image_standard, result_image_hd, idphoto_json
+            result_image_standard, result_image_hd, idphoto_json, language
         )
 
         # 添加水印
@@ -349,7 +357,7 @@ class IDPhotoProcessor:
             idphoto_json,
         )
 
-        # 如果output_image_path_dict为None，
+        # 如果output_image_path_dict为None，即没有设置KB和DPI
         if output_image_path_dict is None:
             return self._create_response(
                 result_image_standard,
@@ -358,9 +366,10 @@ class IDPhotoProcessor:
                 result_image_hd_png,
                 gr.update(value=result_image_layout, visible=result_image_layout_visible),
                 gr.update(value=result_image_template, visible=result_image_template_visible),
+                gr.update(visible = result_image_template_visible),
             )
+        # 如果output_image_path_dict不为None，即设置了KB和DPI
         else:
-            # 如果output_image_path_dict不为None，
             if output_image_path_dict["layout"]["processed"]:
                 result_image_layout = output_image_path_dict["layout"]["path"]
             return self._create_response(
@@ -378,25 +387,40 @@ class IDPhotoProcessor:
                 result_image_hd_png,
                 gr.update(value=result_image_layout, visible=result_image_layout_visible),
                 gr.update(value=result_image_template, visible=result_image_template_visible),
+                gr.update(visible = result_image_template_visible),
             )
 
     # 渲染背景
-    def _render_background(self, result_image_standard, result_image_hd, idphoto_json):
+    def _render_background(self, result_image_standard, result_image_hd, idphoto_json, language):
         """渲染背景"""
         render_modes = {0: "pure_color", 1: "updown_gradient", 2: "center_gradient"}
         render_mode = render_modes[idphoto_json["render_mode"]]
 
-        result_image_standard = np.uint8(
-            add_background(
-                result_image_standard, bgr=idphoto_json["color_bgr"], mode=render_mode
+        if idphoto_json["color_mode"] != LOCALES["bg_color"][language]["choices"][-3]:
+            result_image_standard = np.uint8(
+                add_background(
+                    result_image_standard, bgr=idphoto_json["color_bgr"], mode=render_mode
+                )
             )
-        )
-        result_image_hd = np.uint8(
-            add_background(
-                result_image_hd, bgr=idphoto_json["color_bgr"], mode=render_mode
+            result_image_hd = np.uint8(
+                add_background(
+                    result_image_hd, bgr=idphoto_json["color_bgr"], mode=render_mode
+                )
+            )
+        # 如果选择了美式证件照
+        else:
+            result_image_standard = np.uint8(
+                add_background_with_image(
+                    result_image_standard, 
+                    background_image=cv2.imread(os.path.join(base_path, "assets", "american-style.png"))
+                )
+            )
+            result_image_hd = np.uint8(
+                add_background_with_image(
+                    result_image_hd, 
+                    background_image=cv2.imread(os.path.join(base_path, "assets", "american-style.png"))
+                )
             )
-        )
-
         return result_image_standard, result_image_hd
 
     # 生成排版照片
@@ -425,7 +449,8 @@ class IDPhotoProcessor:
         )
 
         return result_image_layout, True
-
+    
+    # 生成模板照片
     def _generate_image_template(
         self,
         idphoto_json,
@@ -480,6 +505,7 @@ class IDPhotoProcessor:
         result_image_hd,
         result_image_layout,
         idphoto_json,
+        format="png",
     ):
         """如果需要，调整图片大小"""
         # 设置输出路径
@@ -504,10 +530,10 @@ class IDPhotoProcessor:
         if custom_kb and custom_dpi:
             # 为所有输出路径添加DPI信息
             for key in output_paths:
-                output_paths[key]["path"] += f"_{custom_dpi}dpi.jpg"
+                output_paths[key]["path"] += f"_{custom_dpi}dpi.{format}"
             # 为标准图像添加KB信息
             output_paths["standard"]["path"] = output_paths["standard"]["path"].replace(
-                ".jpg", f"_{custom_kb}kb.jpg"
+                f".{format}", f"_{custom_kb}kb.{format}"
             )
 
             # 调整标准图像大小并保存
@@ -534,7 +560,7 @@ class IDPhotoProcessor:
         # 只有自定义DPI的情况
         elif custom_dpi:
             for key in output_paths:
-                output_paths[key]["path"] += f"_{custom_dpi}dpi.jpg"
+                output_paths[key]["path"] += f"_{custom_dpi}dpi.{format}"
                 # 保存所有图像，使用自定义DPI
                 if key == "layout" and result_image_layout is None:
                     pass
@@ -550,7 +576,7 @@ class IDPhotoProcessor:
 
         # 只有自定义KB的情况
         elif custom_kb:
-            output_paths["standard"]["path"] += f"_{custom_kb}kb.jpg"
+            output_paths["standard"]["path"] += f"_{custom_kb}kb.{format}"
             # 只调整标准图像大小并保存
             resize_image_to_kb(
                 result_image_standard,
@@ -573,6 +599,7 @@ class IDPhotoProcessor:
         result_image_hd_png,
         result_layout_image_gr,
         result_image_template_gr,
+        result_image_template_accordion_gr,
     ):
         """创建响应"""
         response = [
@@ -582,6 +609,7 @@ class IDPhotoProcessor:
             result_image_hd_png,
             result_layout_image_gr,
             result_image_template_gr,
+            result_image_template_accordion_gr,
             gr.update(visible=False),
         ]
 

demo/ui.py

@@ -73,6 +73,7 @@ def create_ui(
                 with gr.Tab(
                     LOCALES["key_param"][DEFAULT_LANG]["label"]
                 ) as key_parameter_tab:
+                    # 尺寸模式
                     with gr.Row():
                         mode_options = gr.Radio(
                             choices=LOCALES["size_mode"][DEFAULT_LANG]["choices"],
@@ -80,11 +81,13 @@ def create_ui(
                             value=LOCALES["size_mode"][DEFAULT_LANG]["choices"][0],
                             min_width=520,
                         )
+                        # 人脸对齐
                         face_alignment_options = gr.CheckboxGroup(
                             label=LOCALES["face_alignment"][DEFAULT_LANG]["label"],
                             choices=LOCALES["face_alignment"][DEFAULT_LANG]["choices"],
                             interactive=True,
                         )
+                    # 尺寸列表
                     with gr.Row(visible=True) as size_list_row:
                         size_list_options = gr.Dropdown(
                             choices=LOCALES["size_list"][DEFAULT_LANG]["choices"],
@@ -92,6 +95,7 @@ def create_ui(
                             value=LOCALES["size_list"][DEFAULT_LANG]["choices"][0],
                             elem_id="size_list",
                         )
+                    # 自定义尺寸px
                     with gr.Row(visible=False) as custom_size_px:
                         custom_size_height_px = gr.Number(
                             value=413,
@@ -103,6 +107,7 @@ def create_ui(
                             label=LOCALES["custom_size_px"][DEFAULT_LANG]["width"],
                             interactive=True,
                         )
+                    # 自定义尺寸mm
                     with gr.Row(visible=False) as custom_size_mm:
                         custom_size_height_mm = gr.Number(
                             value=35,
@@ -115,6 +120,7 @@ def create_ui(
                             interactive=True,
                         )
 
+                    # 背景颜色
                     color_options = gr.Radio(
                         choices=LOCALES["bg_color"][DEFAULT_LANG]["choices"],
                         label=LOCALES["bg_color"][DEFAULT_LANG]["label"],
@@ -131,6 +137,7 @@ def create_ui(
                     with gr.Row(visible=False) as custom_color_hex:
                         custom_color_hex_value = gr.Text(value="000000", label="Hex", interactive=True)
 
+                    # 渲染模式
                     render_options = gr.Radio(
                         choices=LOCALES["render_mode"][DEFAULT_LANG]["choices"],
                         label=LOCALES["render_mode"][DEFAULT_LANG]["label"],
@@ -353,19 +360,19 @@ def create_ui(
                     img_output_standard = gr.Image(
                         label=LOCALES["standard_photo"][DEFAULT_LANG]["label"],
                         height=350,
-                        format="jpeg",
+                        format="png",
                     )
                     # 高清照
                     img_output_standard_hd = gr.Image(
                         label=LOCALES["hd_photo"][DEFAULT_LANG]["label"],
                         height=350,
-                        format="jpeg",
+                        format="png",
                     )
                 # 排版照
                 img_output_layout = gr.Image(
                     label=LOCALES["layout_photo"][DEFAULT_LANG]["label"],
                     height=350,
-                    format="jpeg",
+                    format="png",
                 )
                 # 模版照片
                 with gr.Accordion(
@@ -374,7 +381,7 @@ def create_ui(
                     img_output_template = gr.Gallery(
                         label=LOCALES["template_photo"][DEFAULT_LANG]["label"],
                         height=350,
-                        format="jpeg",
+                        format="png",
                     )
                 # 抠图图像
                 with gr.Accordion(
@@ -733,6 +740,7 @@ def create_ui(
                     img_output_standard_hd_png,
                     img_output_layout,
                     img_output_template,
+                    template_image_accordion,
                     notification,
                 ],
             )

hivision/creator/__init__.py

@@ -144,9 +144,10 @@ class IDCreator:
             print("执行人脸对齐")
             print("旋转角度：", ctx.face["roll_angle"])
             # 根据角度旋转原图和抠图
+            b, g, r, a = cv2.split(ctx.matting_image)
             ctx.origin_image, ctx.matting_image, _, _, _, _ = rotate_bound_4channels(
-                ctx.origin_image,
-                cv2.split(ctx.matting_image)[-1],
+                cv2.merge((b, g, r)),
+                a,
                 -1 * ctx.face["roll_angle"],
             )
 

hivision/utils.py

@@ -298,6 +298,36 @@ def add_background(input_image, bgr=(0, 0, 0), mode="pure_color"):
 
     return output
 
+def add_background_with_image(input_image: np.ndarray, background_image: np.ndarray) -> np.ndarray:
+    """
+    本函数的功能为为透明图像加上背景。
+    :param input_image: numpy.array(4 channels), 透明图像
+    :param background_image: numpy.array(3 channels), 背景图像
+    :return: output: 合成好的输出图像
+    """
+    height, width = input_image.shape[:2]
+    try:
+        b, g, r, a = cv2.split(input_image)
+    except ValueError:
+        raise ValueError(
+            "The input image must have 4 channels. 输入图像必须有4个通道，即透明图像。"
+        )
+
+    # 确保背景图像与输入图像大小一致
+    background_image = cv2.resize(background_image, (width, height), cv2.INTER_AREA)
+    background_image = cv2.cvtColor(background_image, cv2.COLOR_BGR2RGB)
+    b2, g2, r2 = cv2.split(background_image)
+
+    a_cal = a / 255.0
+
+    # 修正混合公式
+    output = cv2.merge(
+        (b * a_cal + b2 * (1 - a_cal),
+         g * a_cal + g2 * (1 - a_cal),
+         r * a_cal + r2 * (1 - a_cal))
+    )
+
+    return output.astype(np.uint8)
 
 def add_watermark(
     image, text, size=50, opacity=0.5, angle=45, color="#8B8B1B", space=75