TSTTC
收藏资源简介:
TSTTC数据集是由天津大学和TuSimple共同创建的大型时间到接触(TTC)估计数据集,专注于驾驶场景。该数据集包含超过200,000个序列,覆盖400米的深度范围,每个序列包含六个连续帧,每秒10帧,提供2D和3D边界框以及TTC地面实况。数据集通过数千小时的驾驶数据筛选和平衡不同TTC值的样本,同时使用神经渲染方法生成小TTC案例的片段以增加数据量。TSTTC数据集旨在推动基于单目相机的TTC估计技术,适用于高级驾驶辅助系统和自动驾驶系统,特别是在高速公路上需要更广泛感知能力的场景。
The TSTTC dataset is a large-scale time-to-contact (TTC) estimation dataset jointly created by Tianjin University and TuSimple, focusing on driving scenarios. This dataset contains over 200,000 sequences spanning a depth range of 400 meters. Each sequence consists of six consecutive frames captured at 10 frames per second, and provides 2D and 3D bounding boxes along with TTC ground truth. The dataset is developed by screening thousands of hours of driving data and balancing samples with different TTC values, while also using neural rendering methods to generate segments of cases with small TTC values to expand the dataset size. The TSTTC dataset aims to advance monocular camera-based TTC estimation technologies, and is applicable to Advanced Driver Assistance Systems (ADAS) and autonomous driving systems, particularly in scenarios requiring broader perception capabilities on highways.
- 1TSTTC: A Large-Scale Dataset for Time-to-Contact Estimation in Driving Scenarios天津大学 · 2023年



