jeong2/avoid
收藏资源简介：
- **URL**: https://huggingface.co/datasets/jeong2/avoid - **Dataset url**: http://vi.kaist.ac.kr/avoid/avoid.html - **Supplementary video link**: <!-- http://vi.kaist.ac.kr/avoid/sample.mp4 --> - **Dataset embargo**: We have our project page available, but the full dataset will be open by the end of September. - Since our dataset requires a large amount of storage (>500GB and >1TB with additional data), we expect it will take a couple months to setup a stable server. **Data composition / structure** ``` train / val / test ├──obstacle_weather_Town01_01_01 ├──|──obstacle_weather_Town01_01_01_route0_04_14_20_18_51 ├──|──|──depth: depth map ├──|──|──label_raw: class, extent, position, yaw, num_points, distance, speed, brake, id, ego_matrix ├──|──|──lidar: LiDAR data ├──|──|──lidar_sem: semantic annotations for LiDAR data ├──|──|──measurements: x, y, theta, speed, target_speed, x_command, y_command, command, waypoints, steer, throttle, brake, weather, junction, vehicle_hazard, light_hazard, walker_hazard, stop_sign_hazard, angle, ego_matrix, ├──|──|──rgb: RGB image (left) ├──|──|──rgb_bev: BEV RGB image ├──|──|──rgb_pair: paired RGB image (Right) ├──|──|──semantics: semantic annotations for RGB image (left) ├──|──|──semantics_bev: semantic annotations for BEV RGB image ├──|──|──topdown: encoded topdown view ``` --- ##License: The AVOID dataset is published under the *CC BY-NC-ND 4.0* License, and all codes are published under the *Apache 2.0* License. --- ## Acknowledgement To be updated ## Citation ``` @misc{ avoid, title={avoid}, author={}, year={2023}, url={} } ```
- **URL**：https://huggingface.co/datasets/jeong2/avoid - **数据集官方地址**：http://vi.kaist.ac.kr/avoid/avoid.html - **补充视频链接**：<!-- http://vi.kaist.ac.kr/avoid/sample.mp4 --> - **数据集发布限制**：我们已上线项目页面，但完整数据集将于9月底正式开放。 由于本数据集存储空间需求较大（基础数据超过500GB，含附加数据则超过1TB），我们预计需要数月时间搭建稳定的服务器以提供访问服务。 **数据组成与结构** 训练集 / 验证集 / 测试集 ├──obstacle_weather_Town01_01_01 ├──|──obstacle_weather_Town01_01_01_route0_04_14_20_18_51 ├──|──|──depth: 深度图（depth map） ├──|──|──label_raw: 原始标注文件，包含类别、包围盒范围、位置、偏航角（yaw）、点云数量、距离、车速、制动状态、ID、自车矩阵（ego_matrix） ├──|──|──lidar: 激光雷达（LiDAR）数据 ├──|──|──lidar_sem: 激光雷达语义标注 ├──|──|──measurements: 测量数据，包含x坐标、y坐标、航向角、车速、目标车速、x方向控制指令、y方向控制指令、行驶指令、路点（waypoints）、转向角（steer）、油门开度（throttle）、制动状态、天气状况、路口状态、车辆危险状态、灯光危险状态、行人危险状态、停车标志危险状态、夹角、自车矩阵 ├──|──|──rgb: 左侧RGB图像 ├──|──|──rgb_bev: 鸟瞰视角（BEV）RGB图像 ├──|──|──rgb_pair: 右侧配对RGB图像 ├──|──|──semantics: 左侧RGB图像语义标注 ├──|──|──semantics_bev: 鸟瞰视角RGB图像语义标注 ├──|──|──topdown: 编码式俯视图 --- ## 许可证 本AVOID数据集采用**CC BY-NC-ND 4.0**许可证发布，所有配套代码则采用**Apache 2.0**许可证发布。 --- ## 致谢 待更新 ## 引用 @misc{ avoid, title={avoid}, author={}, year={2023}, url={} }
数据集概述
基本信息
- URL: https://huggingface.co/datasets/jeong2/avoid
- Dataset url: http://vi.kaist.ac.kr/avoid/avoid.html
- License: CC BY-NC-ND 4.0
数据集结构
数据集分为训练集、验证集和测试集，每个子集包含以下文件结构：
train / val / test ├── obstacle_weather_Town01_01_01 ├── |── obstacle_weather_Town01_01_01_route0_04_14_20_18_51 ├── | |── depth: depth map ├── | |── label_raw: class, extent, position, yaw, num_points, distance, speed, brake, id, ego_matrix ├── | |── lidar: LiDAR data ├── | |── lidar_sem: semantic annotations for LiDAR data ├── | |── measurements: x, y, theta, speed, target_speed, x_command, y_command, command, waypoints, steer, throttle, brake, weather, junction, vehicle_hazard, light_hazard, walker_hazard, stop_sign_hazard, angle, ego_matrix ├── | |── rgb: RGB image (left) ├── | |── rgb_bev: BEV RGB image ├── | |── rgb_pair: paired RGB image (Right) ├── | |── semantics: semantic annotations for RGB image (left) ├── | |── semantics_bev: semantic annotations for BEV RGB image ├── | |── topdown: encoded topdown view
数据集访问
- Dataset embargo: 全数据集预计在9月底开放。
- 存储需求: 数据集需要大量存储空间，超过500GB，若包含额外数据则超过1TB。
