Upload 6 files
Browse files由XAI Lab数据集生成引擎提供。包含5000张汽车座舱用户异常动作、行为,包含不同性别、年龄、服饰等特征,动作包含打电话、吸烟、打瞌睡等。
- .gitattributes +1 -0
- csv_to_json_converter.py +136 -0
- dataset_detailed.json +0 -0
- dataset_training.json +0 -0
- demo.csv +6 -0
- pair_5000_prompts_with_details.csv +0 -0
- pair_5000_prompts_with_details.xlsx +3 -0
.gitattributes
CHANGED
|
@@ -57,3 +57,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
|
|
| 57 |
# Video files - compressed
|
| 58 |
*.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 59 |
*.webm filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 57 |
# Video files - compressed
|
| 58 |
*.mp4 filter=lfs diff=lfs merge=lfs -text
|
| 59 |
*.webm filter=lfs diff=lfs merge=lfs -text
|
| 60 |
+
pair_5000_prompts_with_details.xlsx filter=lfs diff=lfs merge=lfs -text
|
csv_to_json_converter.py
ADDED
|
@@ -0,0 +1,136 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import csv
|
| 2 |
+
import json
|
| 3 |
+
import os
|
| 4 |
+
from typing import List, Dict, Any
|
| 5 |
+
|
| 6 |
+
def csv_to_json(csv_file_path: str, output_json_path: str) -> None:
|
| 7 |
+
"""
|
| 8 |
+
将CSV格式的数据集转换为JSON格式用于训练
|
| 9 |
+
|
| 10 |
+
Args:
|
| 11 |
+
csv_file_path: CSV文件路径
|
| 12 |
+
output_json_path: 输出JSON文件路径
|
| 13 |
+
"""
|
| 14 |
+
|
| 15 |
+
data = []
|
| 16 |
+
|
| 17 |
+
with open(csv_file_path, 'r', encoding='utf-8') as csvfile:
|
| 18 |
+
reader = csv.DictReader(csvfile)
|
| 19 |
+
|
| 20 |
+
for row in reader:
|
| 21 |
+
# 构建JSON结构
|
| 22 |
+
json_item = {
|
| 23 |
+
"id": row["id"],
|
| 24 |
+
"prompt_en": row["prompt_en"],
|
| 25 |
+
"image_path": row["path"],
|
| 26 |
+
"persons": [
|
| 27 |
+
{
|
| 28 |
+
"person_id": 1,
|
| 29 |
+
"demographics": row["person1_person"],
|
| 30 |
+
"clothing": {
|
| 31 |
+
"upper_clothing": row["person1_upper_clothing"],
|
| 32 |
+
"upper_clothing_color": row["person1_upper_clothing_color"],
|
| 33 |
+
"lower_body": row["person1_lower_body"],
|
| 34 |
+
"lower_body_color": row["person1_lower_body_color"]
|
| 35 |
+
},
|
| 36 |
+
"accessory": row["person1_accessory"],
|
| 37 |
+
"behavior": row["person1_person_behavior"]
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"person_id": 2,
|
| 41 |
+
"demographics": row["person2_person"],
|
| 42 |
+
"clothing": {
|
| 43 |
+
"upper_clothing": row["person2_upper_clothing"],
|
| 44 |
+
"upper_clothing_color": row["person2_upper_clothing_color"],
|
| 45 |
+
"lower_body": row["person2_lower_body"],
|
| 46 |
+
"lower_body_color": row["person2_lower_body_color"]
|
| 47 |
+
},
|
| 48 |
+
"accessory": row["person2_accessory"],
|
| 49 |
+
"behavior": row["person2_person_behavior"]
|
| 50 |
+
}
|
| 51 |
+
]
|
| 52 |
+
}
|
| 53 |
+
|
| 54 |
+
data.append(json_item)
|
| 55 |
+
|
| 56 |
+
# 保存为JSON文件
|
| 57 |
+
with open(output_json_path, 'w', encoding='utf-8') as jsonfile:
|
| 58 |
+
json.dump(data, jsonfile, ensure_ascii=False, indent=2)
|
| 59 |
+
|
| 60 |
+
print(f"转换完成!共处理 {len(data)} 条数据")
|
| 61 |
+
print(f"JSON文件已保存至: {output_json_path}")
|
| 62 |
+
|
| 63 |
+
def create_training_format(csv_file_path: str, output_json_path: str) -> None:
|
| 64 |
+
"""
|
| 65 |
+
创建适合机器学习训练的简化JSON格式
|
| 66 |
+
|
| 67 |
+
Args:
|
| 68 |
+
csv_file_path: CSV文件路径
|
| 69 |
+
output_json_path: 输出JSON文件路径
|
| 70 |
+
"""
|
| 71 |
+
|
| 72 |
+
training_data = []
|
| 73 |
+
|
| 74 |
+
with open(csv_file_path, 'r', encoding='utf-8') as csvfile:
|
| 75 |
+
reader = csv.DictReader(csvfile)
|
| 76 |
+
|
| 77 |
+
for row in reader:
|
| 78 |
+
# 简化的训练格式
|
| 79 |
+
training_item = {
|
| 80 |
+
"image_id": row["id"],
|
| 81 |
+
"text": row["prompt_en"],
|
| 82 |
+
"image_path": row["path"],
|
| 83 |
+
"labels": {
|
| 84 |
+
"person1": {
|
| 85 |
+
"demographics": row["person1_person"],
|
| 86 |
+
"upper_clothing": row["person1_upper_clothing"],
|
| 87 |
+
"upper_color": row["person1_upper_clothing_color"],
|
| 88 |
+
"lower_clothing": row["person1_lower_body"],
|
| 89 |
+
"lower_color": row["person1_lower_body_color"],
|
| 90 |
+
"accessory": row["person1_accessory"],
|
| 91 |
+
"behavior": row["person1_person_behavior"]
|
| 92 |
+
},
|
| 93 |
+
"person2": {
|
| 94 |
+
"demographics": row["person2_person"],
|
| 95 |
+
"upper_clothing": row["person2_upper_clothing"],
|
| 96 |
+
"upper_color": row["person2_upper_clothing_color"],
|
| 97 |
+
"lower_clothing": row["person2_lower_body"],
|
| 98 |
+
"lower_color": row["person2_lower_body_color"],
|
| 99 |
+
"accessory": row["person2_accessory"],
|
| 100 |
+
"behavior": row["person2_person_behavior"]
|
| 101 |
+
}
|
| 102 |
+
}
|
| 103 |
+
}
|
| 104 |
+
|
| 105 |
+
training_data.append(training_item)
|
| 106 |
+
|
| 107 |
+
# 保存训练格式的JSON
|
| 108 |
+
with open(output_json_path, 'w', encoding='utf-8') as jsonfile:
|
| 109 |
+
json.dump(training_data, jsonfile, ensure_ascii=False, indent=2)
|
| 110 |
+
|
| 111 |
+
print(f"训练格式转换完成!共处理 {len(training_data)} 条数据")
|
| 112 |
+
print(f"训练JSON文件已保存至: {output_json_path}")
|
| 113 |
+
|
| 114 |
+
def main():
|
| 115 |
+
# 设置文件路径
|
| 116 |
+
csv_file = "/Volumes/XAI测试用1号机/Cabin_behavior_dataset/pair_5000_prompts_with_details.csv"
|
| 117 |
+
|
| 118 |
+
# 检查CSV文件是否存在
|
| 119 |
+
if not os.path.exists(csv_file):
|
| 120 |
+
print(f"错误:找不到CSV文件 {csv_file}")
|
| 121 |
+
return
|
| 122 |
+
|
| 123 |
+
# 转换为详细JSON格式
|
| 124 |
+
detailed_json = "/Volumes/XAI测试用1号机/Cabin_behavior_dataset/dataset_detailed.json"
|
| 125 |
+
csv_to_json(csv_file, detailed_json)
|
| 126 |
+
|
| 127 |
+
# 转换为训练JSON格式
|
| 128 |
+
training_json = "/Volumes/XAI测试用1号机/Cabin_behavior_dataset/dataset_training.json"
|
| 129 |
+
create_training_format(csv_file, training_json)
|
| 130 |
+
|
| 131 |
+
print("\n转换完成!生成了两个JSON文件:")
|
| 132 |
+
print(f"1. 详细格式: {detailed_json}")
|
| 133 |
+
print(f"2. 训练格式: {training_json}")
|
| 134 |
+
|
| 135 |
+
if __name__ == "__main__":
|
| 136 |
+
main()
|
dataset_detailed.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
dataset_training.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
demo.csv
ADDED
|
@@ -0,0 +1,6 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
id,prompt_en,person1_person,person1_upper_clothing,person1_upper_clothing_color,person1_lower_body,person1_lower_body_color,person1_accessory,person1_person_behavior,person2_person,person2_upper_clothing,person2_upper_clothing_color,person2_lower_body,person2_lower_body_color,person2_accessory,person2_person_behavior,path
|
| 2 |
+
img_00001,"Top-down view inside a car cockpit, two people as the main subjects: First person: adult female, wearing a sophisticated gray jacket paired with mature brown shorts, vest accessory, playing on a smartphone; Second person: elderly female, wearing a sophisticated gray hoodie paired with vibrant pink tailored trousers, vest accessory, also playing on a smartphone; clean background, realistic style, high definition, natural lighting",成年-女性,夹克,成熟稳重-灰色,短裤,成熟稳重-棕色,马夹,玩手机,老年-女性,连帽衫,成熟稳重-灰色,西装长裤,光鲜靓丽-粉色,马夹,玩手机,/Cabin_behavior_dataset/img_00001.png
|
| 3 |
+
img_00002,"Top-down view inside a car cabin, two people as the main subjects: First person: adult male, wearing a pure and minimalist white coat, paired with mature and sophisticated gray jeans, accessorized with a vest, sleeping; Second person: adult female, wearing a mature and sophisticated brown shirt, paired with vibrant and stylish green casual pants, accessorized with sunglasses, reading; clean background, realistic style, high definition, natural lighting",成年-男性,大衣,纯净简约-白色,牛仔长裤,成熟稳重-灰色,马夹,睡觉,成年-女性,衬衫,成熟稳重-棕色,休闲长裤,光鲜靓丽-绿色,墨镜,阅读,/Cabin_behavior_dataset/img_00002.png
|
| 4 |
+
img_00003,"A top-down view inside a car cockpit, featuring two people as the main subjects: First person: elderly male, wearing a pure minimalist white sweater paired with vibrant orange denim jeans, no accessories, smoking a cigarette; Second person: adult female, wearing a vibrant red hoodie paired with vibrant red tailored trousers, eyewear accessories, applying makeup; clean background, realistic style, high definition, natural lighting",老年-男性,毛衣,纯净简约-白色,牛仔长裤,光鲜靓丽-橙色,无配饰,吸烟,成年-女性,连帽衫,光鲜靓丽-红色,西装长裤,光鲜靓丽-红色,眼镜,化妆,/Cabin_behavior_dataset/img_00003.png
|
| 5 |
+
img_00004,"Top-down view inside a car cockpit, two people as the main subjects: First person: adult female, wearing a mature and dignified brown T-shirt paired with vibrant green straight-leg jeans, accessorized with a scarf, playing a smartphone; Second person: adult male, wearing a mature and elegant brown coat paired with sophisticated gray casual trousers, wearing a mask, smoking; clean background, realistic style, high definition, natural lighting",成年-女性,T恤,成熟稳重-棕色,牛仔长裤,光鲜靓丽-绿色,围巾,玩手机,成年-男性,大衣,成熟稳重-棕色,休闲长裤,成熟稳重-灰色,口罩,吸烟,/Cabin_behavior_dataset/img_00004.png
|
| 6 |
+
img_00005,"Top-down view inside a car cockpit, two elderly women as main subjects: First person: elderly female, wearing a vibrant pink jacket paired with a sophisticated purple short, no accessories, sleeping peacefully; Second person: elderly female, wearing a bright orange shirt paired with vibrant yellow tailored trousers, with a scarf as accessory, eating; clean background, realistic style, high definition, natural lighting",老年-女性,夹克,光鲜靓丽-粉色,短裤,成熟稳重-紫色,无配饰,睡觉,老年-女性,衬衫,光鲜靓丽-橙色,西装长裤,光鲜靓丽-黄色,围巾,吃东西,/Cabin_behavior_dataset/img_00005.png
|
pair_5000_prompts_with_details.csv
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
pair_5000_prompts_with_details.xlsx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a965d71e78d1e18fc88d6b5008776c927ee6a6bbd4fdb445d349ea101b6ca9cf
|
| 3 |
+
size 662768
|