Add files using upload-large-folder tool

.gitattributes

@@ -0,0 +1,42 @@
+pics/10_img_00002.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00005.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00001.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00006.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00004.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00003.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00008.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00010.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00007.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00009.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00014.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00011.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00018.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00013.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00012.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00015.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00016.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00017.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00019.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00021.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00020.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00022.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00023.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00024.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00025.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00027.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00026.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00029.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00028.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00030.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00031.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00032.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00033.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00035.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00034.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00036.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00037.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00038.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00039.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00040.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00042.png filter=lfs diff=lfs merge=lfs -text
+pics/10_img_00041.png filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,49 @@
+
+# 汽车座舱人类异常行为数据集 (Cabin Human ABNORMAL Behavior Dataset)
+
+## 数据集描述
+
+此数据集包含汽车座舱中左右乘客的异常行为图像及其标签。数据集分为训练集和测试集，以便于模型训练和评估。
+
+## 数据格式
+
+数据集以JSON格式提供，包含以下字段：
+
+- `image_id`: 图像ID
+- `image_path`: 图像路径
+- `category`: 行为类别
+- `tags`: 行为标签
+- `behaviors`: 包含左右乘客行为描述的对象
+  - `left_passenger`: 左侧乘客行为描述
+  - `right_passenger`: 右侧乘客行为描述
+
+## 数据集统计
+
+- 总数据量：4300条
+- 训练集：3440条
+- 测试集：860条
+- 类别数：43个
+
+## 使用方法
+
+```python
+from datasets import load_dataset
+
+# 加载数据集
+dataset = load_dataset("XAILab-CyberSpark/Cabin-Human-ABNORMAL-Behavior-Dataset")
+
+# 查看数据集信息
+print(dataset)
+```
+
+## 引用
+
+如果您在研究中使用了此数据集，请引用：
+
+```
+@dataset{cabin_human_abnormal_behavior_dataset,
+  author       = {XAILab-CyberSpark},
+  title        = {Cabin Human ABNORMAL Behavior Dataset},
+  year         = {2024},
+}
+```

generate_json_tags.py

@@ -0,0 +1,304 @@
+import csv
+import json
+import os
+import random
+from typing import List, Dict, Any, Tuple
+
+def get_image_files(pics_dir: str) -> Dict[int, List[str]]:
+    """
+    遍历图片目录，按类别整理图片文件名
+    
+    Args:
+        pics_dir: 图片目录路径
+        
+    Returns:
+        按类别分组的图片文件名字典
+    """
+    image_files = {}
+    
+    # 确保目录存在
+    if not os.path.exists(pics_dir):
+        print(f"错误：图片目录 {pics_dir} 不存在")
+        return {}
+    
+    # 遍历目录中的所有文件
+    for filename in os.listdir(pics_dir):
+        if filename.endswith('.png'):
+            # 解析文件名，例如：1_img_00001.png
+            parts = filename.split('_')
+            if len(parts) >= 3:
+                try:
+                    category_id = int(parts[0])
+                    if category_id not in image_files:
+                        image_files[category_id] = []
+                    image_files[category_id].append(filename)
+                except ValueError:
+                    continue
+    
+    # 对每个类别的图片列表进行排序
+    for category_id in image_files:
+        image_files[category_id].sort()
+    
+    print(f"找到 {len(image_files)} 个类别的图片")
+    for category_id, files in image_files.items():
+        print(f"类别 {category_id}: {len(files)} 张图片")
+    
+    return image_files
+
+def split_train_test(image_files: Dict[int, List[str]], test_ratio: float = 0.2) -> Tuple[Dict[int, List[str]], Dict[int, List[str]]]:
+    """
+    将图片分割为训练集和测试集
+    
+    Args:
+        image_files: 按类别分组的图片文件名字典
+        test_ratio: 测试集比例
+        
+    Returns:
+        训练集和测试集的图片文件名字典
+    """
+    train_files = {}
+    test_files = {}
+    
+    for category_id, files in image_files.items():
+        # 随机打乱文件列表
+        random.shuffle(files)
+        
+        # 计算测试集大小
+        test_size = max(1, int(len(files) * test_ratio))
+        
+        # 分割训练集和测试集
+        test_files[category_id] = files[:test_size]
+        train_files[category_id] = files[test_size:]
+    
+    return train_files, test_files
+
+def create_json_dataset(csv_file_path: str, pics_dir: str, output_train_json: str, output_test_json: str, test_ratio: float = 0.2) -> None:
+    """
+    创建训练集和测试集的JSON标签文件
+    
+    Args:
+        csv_file_path: CSV文件路径
+        pics_dir: 图片目录路径
+        output_train_json: 输出训练集JSON文件路径
+        output_test_json: 输出测试集JSON文件路径
+        test_ratio: 测试集比例
+    """
+    # 获取图片文件
+    image_files = get_image_files(pics_dir)
+    if not image_files:
+        print("未找到有效的图片文件")
+        return
+    
+    # 读取CSV文件
+    csv_data = {}
+    with open(csv_file_path, 'r', encoding='utf-8') as csvfile:
+        reader = csv.DictReader(csvfile)
+        for row in reader:
+            try:
+                num = int(row['num'])
+                csv_data[num] = row
+            except (ValueError, KeyError):
+                continue
+    
+    print(f"从CSV文件中读取了 {len(csv_data)} 个类别的数据")
+    
+    # 分割训练集和测试集
+    train_files, test_files = split_train_test(image_files, test_ratio)
+    
+    # 创建训练集和测试集数据
+    train_data = []
+    test_data = []
+    
+    # 处理训练集
+    for category_id, files in train_files.items():
+        if category_id not in csv_data:
+            print(f"警告：类别 {category_id} 在CSV文件中不存在")
+            continue
+            
+        category_info = csv_data[category_id]
+        
+        for image_file in files:
+            json_item = {
+                "image_id": image_file.split('.')[0],
+                "image_path": os.path.join(pics_dir, image_file),
+                "category": category_info.get('category', ''),
+                "tags": category_info.get('tags', ''),
+                "behaviors": {
+                    "left_passenger": category_info.get('describe_left', ''),
+                    "right_passenger": category_info.get('describe_right', '')
+                }
+            }
+            train_data.append(json_item)
+    
+    # 处理测试集
+    for category_id, files in test_files.items():
+        if category_id not in csv_data:
+            print(f"警告：类别 {category_id} 在CSV文件中不存在")
+            continue
+            
+        category_info = csv_data[category_id]
+        
+        for image_file in files:
+            json_item = {
+                "image_id": image_file.split('.')[0],
+                "image_path": os.path.join(pics_dir, image_file),
+                "category": category_info.get('category', ''),
+                "tags": category_info.get('tags', ''),
+                "behaviors": {
+                    "left_passenger": category_info.get('describe_left', ''),
+                    "right_passenger": category_info.get('describe_right', '')
+                }
+            }
+            test_data.append(json_item)
+    
+    # 保存训练集JSON
+    with open(output_train_json, 'w', encoding='utf-8') as jsonfile:
+        json.dump(train_data, jsonfile, ensure_ascii=False, indent=2)
+    
+    # 保存测试集JSON
+    with open(output_test_json, 'w', encoding='utf-8') as jsonfile:
+        json.dump(test_data, jsonfile, ensure_ascii=False, indent=2)
+    
+    print(f"转换完成！")
+    print(f"训练集：{len(train_data)} 条数据，已保存至: {output_train_json}")
+    print(f"测试集：{len(test_data)} 条数据，已保存至: {output_test_json}")
+
+def create_direct_json_dataset(csv_file_path: str, pics_dir: str, output_json_path: str) -> None:
+    """
+    直接创建JSON标签文件，不分训练集和测试集
+    
+    Args:
+        csv_file_path: CSV文件路径
+        pics_dir: 图片目录路径
+        output_json_path: 输出JSON文件路径
+    """
+    # 读取CSV文件
+    csv_data = {}
+    
+    # 直接使用pandas读取CSV文件，处理BOM标记
+    import pandas as pd
+    try:
+        df = pd.read_csv(csv_file_path)
+        # 检查列名是否有BOM标记
+        if 'num' in df.columns:
+            df = df.rename(columns={'num': 'num'})
+        
+        # 将DataFrame转换为字典
+        for _, row in df.iterrows():
+            try:
+                num = int(row['num'])
+                csv_data[num] = row.to_dict()
+            except (ValueError, KeyError) as e:
+                print(f"处理行时出错: {e}")
+                continue
+    except Exception as e:
+        print(f"使用pandas读取CSV文件时出错: {e}")
+    
+    print(f"从CSV文件中读取了 {len(csv_data)} 个类别的数据")
+    if len(csv_data) > 0:
+        print(f"CSV数据示例: {list(csv_data.keys())[:5]}")
+    
+    # 获取所有图片文件
+    all_images = []
+    for filename in os.listdir(pics_dir):
+        if filename.endswith('.png'):
+            all_images.append(filename)
+    
+    print(f"找到 {len(all_images)} 张图片")
+    if len(all_images) > 0:
+        print(f"图片文件示例: {all_images[:5]}")
+    
+    # 创建JSON数据
+    json_data = []
+    
+    for image_file in all_images:
+        # 解析文件名，例如：1_img_00001.png
+        parts = image_file.split('_')
+        if len(parts) >= 3:
+            try:
+                category_id = int(parts[0])
+                if category_id in csv_data:
+                    category_info = csv_data[category_id]
+                    
+                    json_item = {
+                        "image_id": image_file.split('.')[0],
+                        "image_path": os.path.join(pics_dir, image_file),
+                        "category": category_info.get('category', ''),
+                        "tags": category_info.get('tags', ''),
+                        "behaviors": {
+                            "left_passenger": category_info.get('describe_left', ''),
+                            "right_passenger": category_info.get('describe_right', '')
+                        }
+                    }
+                    json_data.append(json_item)
+            except ValueError:
+                continue
+    
+    # 保存JSON文件
+    with open(output_json_path, 'w', encoding='utf-8') as jsonfile:
+        json.dump(json_data, jsonfile, ensure_ascii=False, indent=2)
+    
+    print(f"转换完成！共生成 {len(json_data)} 条数据，已保存至: {output_json_path}")
+
+def main():
+    """主函数"""
+    # 设置文件路径
+    csv_file_path = "/Volumes/XAI测试二号机/to_zwj/tags_for_gen.csv"
+    pics_dir = "/Volumes/XAI测试二号机/to_zwj/pics"
+    output_json_path = "/Volumes/XAI测试二号机/to_zwj/dataset_all.json"
+    
+    # 检查CSV文件是否存在
+    if not os.path.exists(csv_file_path):
+        print(f"错误：找不到CSV文件 {csv_file_path}")
+        return
+    
+    # 检查图片目录是否存在
+    if not os.path.exists(pics_dir):
+        print(f"错误：找不到图片目录 {pics_dir}")
+        return
+    
+    # 直接创建JSON标签文件，不分训练集和测试集
+    create_direct_json_dataset(csv_file_path, pics_dir, output_json_path)
+    
+    # 从生成的JSON文件中分割训练集和测试集
+    train_json_path = "/Volumes/XAI测试二号机/to_zwj/dataset_train.json"
+    test_json_path = "/Volumes/XAI测试二号机/to_zwj/dataset_test.json"
+    
+    # 读取生成的JSON文件
+    with open(output_json_path, 'r', encoding='utf-8') as f:
+        all_data = json.load(f)
+    
+    # 按类别分组
+    data_by_category = {}
+    for item in all_data:
+        category = item['category']
+        if category not in data_by_category:
+            data_by_category[category] = []
+        data_by_category[category].append(item)
+    
+    # 分割训练集和测试集
+    train_data = []
+    test_data = []
+    test_ratio = 0.2
+    
+    for category, items in data_by_category.items():
+        # 随机打乱
+        random.shuffle(items)
+        # 计算测试集大小
+        test_size = max(1, int(len(items) * test_ratio))
+        # 分割
+        test_data.extend(items[:test_size])
+        train_data.extend(items[test_size:])
+    
+    # 保存训练集
+    with open(train_json_path, 'w', encoding='utf-8') as f:
+        json.dump(train_data, f, ensure_ascii=False, indent=2)
+    
+    # 保存测试集
+    with open(test_json_path, 'w', encoding='utf-8') as f:
+        json.dump(test_data, f, ensure_ascii=False, indent=2)
+    
+    print(f"数据集分割完成！训练集: {len(train_data)} 条数据，测试集: {len(test_data)} 条数据")
+
+if __name__ == "__main__":
+    main()

tags_for_gen.csv

@@ -0,0 +1,44 @@
+num,category,tags,describe_left,describe_right
+1,心血管系统紧急状况,手捂胸口 (左/中),单手紧紧捂住胸口，神情痛苦,表情烦躁，整理坐姿
+2,心血管系统紧急状况,手抓握左臂,手抓握左臂，神情痛苦,表情疑惑，望向右边
+3,心血管系统紧急状况,手触碰下颚/颈部,手触碰下颚，神情痛苦,单侧面部下垂，难过
+4,心血管系统紧急状况,身体前倾/蜷缩,身体轻微蜷缩，面部痛苦,单手按压肚子，表情痛苦
+5,心血管系统紧急状况,身体僵直靠椅,身体僵直靠椅，闭眼皱眉，神情虚弱,单手紧紧捂住胸口，神情痛苦
+6,心血管系统紧急状况,面部痛苦扭曲,面部肌肉扭曲，眉头紧锁，咬紧牙关。,单手揉搓眼睛
+7,心血管系统紧急状况,面色苍白,面色微白，表情虚弱，闭眼。,的面色稍白，嘴唇失去血气，表情虚弱
+8,心血管系统紧急状况,嘴唇发绀,面色稍白，嘴唇失去血气，表情虚弱,低头，寻找哮喘喷雾
+9,心血管系统紧急状况,大汗淋漓/冒冷汗,流汗,感到寒冷，双手抱紧自己，闭眼，皱眉
+10,神经系统紧急状况,单侧面部下垂,单侧面部下垂，难过,呼吸极度费力，鼻翼随呼吸张合。
+11,神经系统紧急状况,口角歪斜,试图说话或微笑时，嘴巴歪向一侧。,昏倒
+12,神经系统紧急状况,手臂无力下垂,一只手臂无力滑落或下垂,看起来湿透了，表情虚弱
+13,神经系统紧急状况,眼神呆滞/凝视,眼睛望向前方，嘴巴张开，发呆,流汗
+14,神经系统紧急状况,身体强直/僵硬,身体僵硬，神态恐慌,眯眼，用手遮挡来自窗外的光线
+15,神经系统紧急状况,身体节律性抽搐,向一边倒，面无表情，闭眼，头向一左边转,面部肌肉扭曲，眉头紧锁，咬紧牙关。
+16,神经系统紧急状况,双眼上翻,眼球上翻，嘴巴张大,面部肿胀
+17,神经系统紧急状况,口吐白沫,嘴边出现唾液泡沫，神情难受，闭眼,面色微白，表情虚弱，闭眼。
+18,呼吸系统紧急状况,手掐喉咙/颈部,双手紧紧抓住或掐住自己的喉咙，神情难受，闭眼,忍受的剧痛，双手保护身体
+19,呼吸系统紧急状况,张口大口喘气,嘴巴张得很大，双手紧紧抓住或掐住自己的喉咙，颈部肌肉紧张，费力地呼吸。,身体僵硬，神态恐慌
+20,呼吸系统紧急状况,身体前倾辅助呼吸,身体前倾，嘴巴张大，双手支撑在膝盖上，呼吸困难，试图让肺部扩张。,身体僵直靠椅，闭眼皱眉，神情虚弱
+21,呼吸系统紧急状况,鼻翼扇动,呼吸极度费力，鼻翼随呼吸张合。,身体前倾，嘴巴张大，双手支撑在膝盖上，呼吸困难，试图让肺部扩张。
+22,呼吸系统紧急状况,寻找吸入器/药物,低头，寻找哮喘喷雾,身体前倾，嘴巴张大吐出脏水，闭眼，皱眉
+23,呼吸系统紧急状况,面部/眼睑肿胀,面部肿胀,身体轻微蜷缩，面部痛苦
+24,消化系统紧急状况,手捂腹部,单手按压肚子，表情痛苦,试图说话或微笑时，嘴巴歪向一侧。
+25,消化系统紧急状况,身体蜷缩/弯腰,向前弯腰，腹部剧痛，表情痛苦,手触碰下颚，神情痛苦
+26,消化系统紧急状况,干呕/作呕迹象,嘴巴嘟起微微伸出舌头，闭眼，皱眉,手伏在车门上，神情恐慌
+27,消化系统紧急状况,手捂嘴准备呕吐,双手迅速捂住嘴巴，闭眼，皱眉,手举在手上，打哈欠
+28,消化系统紧急状况,正在呕吐,身体前倾，嘴巴张大吐出脏水，闭眼，皱眉,手抓握左臂，神情痛苦
+29,代谢系统紧急状况,身体无法控制地颤抖,感到寒冷，双手抱紧自己，闭眼，皱眉,双手抱住头部，表情痛苦，闭眼
+30,代谢系统紧急状况,异常大量出汗,看起来湿透了，表情虚弱,双手紧紧抓住或掐住自己的喉咙，神情难受，闭眼
+31,代谢系统紧急状况,寻找食物/糖分,在车内（如储物格）焦急地翻找东西，可能是寻找糖果或含糖饮料。,双手食指用力按压两侧太阳穴，表情痛苦，闭眼
+32,一般性疼痛与不适,手捂头部,双手抱住头部，表情痛苦，闭眼,双手迅速捂住嘴巴，闭眼，皱眉
+33,一般性疼痛与不适,双手按压太阳穴,双手食指用力按压两侧太阳穴，表情痛苦，闭眼,头部垂下，看到头顶
+34,一般性疼痛与不适,手揉眼睛/前额,单手揉搓眼睛,向前弯腰，腹部剧痛，表情痛苦
+35,一般性疼痛与不适,畏光/遮挡光线,眯眼，用手遮挡来自窗外的光线,向一边倒，面无表情，闭眼，头向一左边转
+36,一般性疼痛与不适,身体烦躁不安,表情烦躁，整理坐姿,眼睛望向前方，嘴巴张开，发呆
+37,一般性疼痛与不适,身体异常扭曲,忍受的剧痛，双手保护身体,眼球翻白，失去瞳孔，嘴巴张大
+38,意识水平改变,头部无力下垂/摇晃,头部垂下，看到头顶,眼神空洞
+39,意识水平改变,眼神涣散/失焦,眼神空洞,一只手臂无力滑落或下垂
+40,意识水平改变,反应迟钝/无反应,手伏在车门上，神情恐慌,在车内（如储物格）焦急地翻找东西，可能是寻找糖果或含糖饮料。
+41,意识水平改变,完全昏厥/瘫倒,昏倒,嘴巴嘟起微微伸出舌头，闭眼，皱眉
+42,意识水平改变,意识模糊/迷茫,表情疑惑，望向右边,嘴巴张得很大，双手紧紧抓住自己的喉咙
+43,意识水平改变,反复打哈欠,手举在手上，打哈欠,嘴边出现唾液泡沫，神情难受，闭眼