Upload 2 files

freeway_demo_ema_model/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:489eed99d7bf7f8ecfdd933c7f0017efab6678942b7f8187091fc379a32018bc
+size 6066952715

script/process_images.py

@@ -0,0 +1,127 @@
+# -*- coding: utf-8 -*-
+
+import os
+from PIL import Image
+import numpy as np
+import json
+
+Image.MAX_IMAGE_PIXELS = None
+
+from concurrent.futures import ThreadPoolExecutor
+from tqdm import tqdm
+
+max_pixels=2048*2048
+ 
+max_long_size=4096
+def has_alpha(img:Image.Image):
+    for band in img.getbands():
+        if band in {'A','a','P'}:
+            return True
+    return False
+
+def add_white_background(img:Image.Image)->Image.Image:
+    img=img.convert('RGBA') #转换为RGBA
+    background = Image.new('RGBA', img.size, (255, 255, 255)) 
+    img = Image.alpha_composite(background, img) 
+    return img
+
+def resize_image(image:Image.Image)->Image.Image:
+ 
+    width, height = image.size
+    max_side = max(width, height)
+    current_pixels=width*height
+ 
+    # 检查是否需要调整大小
+    if max_side > max_long_size or current_pixels>max_pixels:
+ 
+        # 计算缩放比例
+        scale = min((max_long_size / max_side),
+                    ((max_pixels / current_pixels) ** 0.5))
+        # 计算新的尺寸
+        new_width = int(width * scale)
+        new_height = int(height * scale)
+        # 调整图片大小
+        resized_image = image.resize((new_width, new_height),
+                                      Image.BICUBIC
+                                      )
+        return resized_image
+    # 如果不需要调整大小，返回原始图片
+    return image
+ 
+def load_image(image_path:str)->Image.Image:
+    try:
+        with Image.open(image_path) as img:
+            img.load()#读取图片加载到内存
+            np.array(img) #尝试用numpy加载图片
+            img=resize_image(img) #resize图片
+            if has_alpha(img): #读取并移除透明图层
+                img=add_white_background(img) #添加白色背景
+            if not img.mode == "RGB":
+                img = img.convert("RGB")
+            return img
+    except:
+        return None
+
+def get_image_metainfo(img):
+    if img is None:
+        return None
+    else:
+        width, height = img.size
+        return {'width':width,
+                'height':height,
+                'pixel_num':width*height,
+                
+            }
+    
+    
+def process_image(input_image_path:str,output_image_path:str):
+
+    img=load_image(input_image_path)
+    
+    image_metainfo=get_image_metainfo(img)
+    
+    output_image_json_path=output_image_path.replace(".webp",".json")
+    
+    
+    if img is not None and image_metainfo is not None:
+        img.save(output_image_path,"WEBP",quality=90) #保存图像
+        with open(output_image_json_path,'w') as f: #保存metainfo
+            json.dump(image_metainfo,f,indent=4)
+            
+def get_image_paths(input_dir, output_dir):
+    for root, _, files in os.walk(input_dir):
+        for file in files:
+            if file.lower().endswith(('.png', '.jpg', '.jpeg', '.webp', '.bmp')):
+                input_path = os.path.join(root, file)
+                rel_path = os.path.relpath(input_path,
+                                           input_dir)
+                output_path = os.path.join(output_dir,
+                                           os.path.splitext(rel_path)[0] + '.webp')
+                os.makedirs(os.path.dirname(output_path), exist_ok=True)
+                yield input_path, output_path
+                
+def process_images_with_thread_pool(input_image_dir:str,
+                                    output_image_dir:str,
+                                    num_threads=16):
+    os.makedirs(output_image_dir, exist_ok=True) 
+    image_paths = get_image_paths(input_image_dir, output_image_dir)
+    with ThreadPoolExecutor(max_workers=num_threads) as executor:
+        # 创建任务列表
+        futures = []
+        for input_path, output_path in image_paths:
+            futures.append(executor.submit(process_image,
+                                           input_path,
+                                           output_path))
+        for _ in tqdm(
+                executor.map(lambda f: f.result(), futures),
+                total=len(futures),
+                desc="Processing images"):
+            pass
+         
+if __name__ == "__main__":
+    # process_image(
+    #     input_image_path="test.png",
+    #     output_image_path='test.webp')
+    process_images_with_thread_pool(input_image_dir=r"20240808\unsplash-research-dataset-lite-latest\test",
+                                        output_image_dir=r"20240808\unsplash-research-dataset-lite-latest\output",
+                                        num_threads=16)