Update app.py

app.py

@@ -148,15 +148,8 @@ def get_random_bool():
     return random.choice([True, False])
 
 def add_white_border(input_image, border_width=10):
-    """
-    为PIL图像添加指定宽度的白色边框。
-    
-    :param input_image: PIL图像对象
-    :param border_width: 边框宽度（单位：像素）
-    :return: 带有白色边框的PIL图像对象
-    """
-    border_color = 'white'  # 白色边框
-    # 添加边框
+
+    border_color = 'white'  
     img_with_border = ImageOps.expand(input_image, border=border_width, fill=border_color)
     return img_with_border
 
@@ -406,26 +399,20 @@ VISION_CONFIG_DICT = {
 }
 
 def concat_images_vertically_and_scale(images,scale_factor=2):
-    # 加载所有图像
-    # 确保所有图像的宽度一致
     widths = [img.width for img in images]
     if not all(width == widths[0] for width in widths):
         raise ValueError('All images must have the same width.')
-    
-    # 计算总高度
+ 
     total_height = sum(img.height for img in images)
     
-    # 创建新的图像，宽度与原图相同，高度为所有图像高度之和
     max_width = max(widths)
     concatenated_image = Image.new('RGB', (max_width, total_height))
 
-    # 竖直拼接图像
     current_height = 0
     for img in images:
         concatenated_image.paste(img, (0, current_height))
         current_height += img.height
 
-    # 缩放图像为1/n高度
     new_height = concatenated_image.height // scale_factor
     new_width = concatenated_image.width // scale_factor
     resized_image = concatenated_image.resize((new_width, new_height), Image.ANTIALIAS)
@@ -434,19 +421,13 @@ def concat_images_vertically_and_scale(images,scale_factor=2):
 
 
 def combine_images_horizontally(images):
-    # 读取所有图片并存入列表
-
-    # 获取每幅图像的宽度和高度
     widths, heights = zip(*(i.size for i in images))
 
-    # 计算总宽度和最大高度
     total_width = sum(widths)
     max_height = max(heights)
 
-    # 创建新的空白图片，用于拼接
     new_im = Image.new('RGB', (total_width, max_height))
 
-    # 将图片横向拼接
     x_offset = 0
     for im in images:
         new_im.paste(im, (x_offset, 0))
@@ -456,28 +437,20 @@ def combine_images_horizontally(images):
 
 def combine_images_vertically_with_resize(images):
     
-    # 获取所有图片的宽度和高度
     widths, heights = zip(*(i.size for i in images))
     
-    # 确定新图片的宽度，即所有图片中最小的宽度
     min_width = min(widths)
     
-    # 调整图片尺寸以保持宽度一致，长宽比不变
     resized_images = []
     for img in images:
-        # 计算新高度保持图片长宽比
         new_height = int(min_width * img.height / img.width)
-        # 调整图片大小
         resized_img = img.resize((min_width, new_height), Image.ANTIALIAS)
         resized_images.append(resized_img)
     
-    # 计算所有调整尺寸后图片的总高度
     total_height = sum(img.height for img in resized_images)
     
-    # 创建一个足够宽和高的新图片对象
     new_im = Image.new('RGB', (min_width, total_height))
     
-    # 竖直拼接图片
     y_offset = 0
     for im in resized_images:
         new_im.paste(im, (0, y_offset))
@@ -516,10 +489,7 @@ def distribute_images(images, group_sizes=(4, 3, 2)):
     remaining = len(images)
     
     while remaining > 0:
-        # 优先分配最大组（4张图片），再考虑3张，最后处理2张
         for size in sorted(group_sizes, reverse=True):
-            # 如果剩下的图片数量大于等于当前组大小，或者为图片总数时（也就是第一次迭代）
-            # 开始创建新组
             if remaining >= size or remaining == len(images):
                 if remaining > size:
                     new_group = images[-remaining: -remaining + size]
@@ -528,7 +498,6 @@ def distribute_images(images, group_sizes=(4, 3, 2)):
                 groups.append(new_group)
                 remaining -= size
                 break
-            # 如果剩下的图片少于最小的组大小（2张）并且已经有组了，就把剩下的图片加到最后一个组
             elif remaining < min(group_sizes) and groups:
                 groups[-1].extend(images[-remaining:])
                 remaining = 0
@@ -637,35 +606,15 @@ style_list = [
         "prompt": "{prompt}",
         "negative_prompt": "",
     },
-    {
-        "name": "Japanese Anime",
-        "prompt": "anime artwork illustrating {prompt}. created by japanese anime studio. highly emotional. best quality, high resolution",
-        "negative_prompt": "low quality, low resolution"
-    },
     {
         "name": "Cinematic",
         "prompt": "cinematic still {prompt} . emotional, harmonious, vignette, highly detailed, high budget, bokeh, cinemascope, moody, epic, gorgeous, film grain, grainy",
         "negative_prompt": "anime, cartoon, graphic, text, painting, crayon, graphite, abstract, glitch, deformed, mutated, ugly, disfigured",
     },
-    {
-        "name": "Disney Charactor",
-        "prompt": "A Pixar animation character of {prompt} . pixar-style, studio anime, Disney, high-quality",
-        "negative_prompt": "lowres, bad anatomy, bad hands, text, bad eyes, bad arms, bad legs, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, blurry, grayscale, noisy, sloppy, messy, grainy, highly detailed, ultra textured, photo",
-    },
     {
         "name": "Photographic",
         "prompt": "cinematic photo {prompt} . 35mm photograph, film, bokeh, professional, 4k, highly detailed",
         "negative_prompt": "drawing, painting, crayon, sketch, graphite, impressionist, noisy, blurry, soft, deformed, ugly",
-    },
-    {
-        "name": "Comic book",
-        "prompt": "comic {prompt} . graphic illustration, comic art, graphic novel art, vibrant, highly detailed",
-        "negative_prompt": "photograph, deformed, glitch, noisy, realistic, stock photo",
-    },
-    {
-        "name": "Line art",
-        "prompt": "line art drawing {prompt} . professional, sleek, modern, minimalist, graphic, line art, vector graphics",
-        "negative_prompt": "anime, photorealistic, 35mm film, deformed, glitch, blurry, noisy, off-center, deformed, cross-eyed, closed eyes, bad anatomy, ugly, disfigured, mutated, realism, realistic, impressionism, expressionism, oil, acrylic",
     }
 ]
 
@@ -675,7 +624,7 @@ image_encoder_path = "./data/models/ip_adapter/sdxl_models/image_encoder"
 ip_ckpt = "./data/models/ip_adapter/sdxl_models/ip-adapter_sdxl_vit-h.bin"
 os.environ["no_proxy"] = "localhost,127.0.0.1,::1"
 STYLE_NAMES = list(styles.keys())
-DEFAULT_STYLE_NAME = "Japanese Anime"
+DEFAULT_STYLE_NAME = "Cinematic"
 global models_dict
 use_va = True
 models_dict = {
@@ -748,7 +697,6 @@ class SpatialAttnProcessor2_0(torch.nn.Module):
             self.id_bank[cur_step] = [hidden_states[:self.id_length], hidden_states[self.id_length:]]
         else:
             encoder_hidden_states = torch.cat((self.id_bank[cur_step][0].to(self.device),hidden_states[:1],self.id_bank[cur_step][1].to(self.device),hidden_states[1:]))
-        # 判断随机数是否大于0.5
         if cur_step <=1:
             hidden_states = self.__call2__(attn, hidden_states,None,attention_mask,temb)
         else:   # 256 1024 4096
@@ -1001,39 +949,16 @@ css = '''
 
 #################################################
 title = r"""
-<h1 align="center">StoryDiffusion: Consistent Self-Attention for Long-Range Image and Video Generation</h1>
+<h1 align="center">Demo for Consistent Self-Attention for Long-Range Image and Video Generation</h1>
 """
 
 description = r"""
-<b>Official 🤗 Gradio demo</b> for <a href='https://github.com/HVision-NKU/StoryDiffusion' target='_blank'><b>StoryDiffusion: Consistent Self-Attention for Long-Range Image and Video Generation</b></a>.<br>
-❗️❗️❗️[<b>Important</b>] Personalization steps:<br>
 1️⃣ Enter a Textual Description for Character, if you add the Ref-Image, making sure to <b>follow the class word</b> you want to customize with the <b>trigger word</b>: `img`, such as: `man img` or `woman img` or `girl img`.<br>
 2️⃣ Enter the prompt array, each line corrsponds to one generated image.<br>
 3️⃣ Choose your preferred style template.<br>
 4️⃣ Click the <b>Submit</b> button to start customizing.
 """
 
-article = r"""
-If StoryDiffusion is helpful, please help to ⭐ the <a href='https://github.com/HVision-NKU/StoryDiffusion' target='_blank'>Github Repo</a>. Thanks! 
-[![GitHub Stars](https://img.shields.io/github/stars/HVision-NKU/StoryDiffusion?style=social)](https://github.com/HVision-NKU/StoryDiffusion)
----
-📝 **Citation**
-<br>
-If our work is useful for your research, please consider citing:
-```bibtex
-@article{Zhou2024storydiffusion,
-  title={StoryDiffusion: Consistent Self-Attention for Long-Range Image and Video Generation},
-  author={Zhou, Yupeng and Zhou, Daquan and Cheng, Ming-Ming and Feng, Jiashi and Hou, Qibin},
-  year={2024}
-}
-```
-📋 **License**
-<br>
-The Contents you create are under Apache-2.0 LICENSE. The Code are under Attribution-NonCommercial 4.0 International. 
-📧 **Contact**
-<br>
-If you have any questions, please feel free to reach me out at <b>ypzhousdu@gmail.com</b>.
-"""
 version = r"""
 <h3 align="center">StoryDiffusion Version 0.01 (test version)</h3>
 <h5 >1. Support image ref image. (Cartoon Ref image is not support now)</h5>