This is the official repository for TEPR.
Edge-only reconstruction results observed from different viewpoints. The aligned edge structures across views indicate that edge-only input can still produce a reasonable reconstruction.
Sample ROS bag files and processed data are available on Google Drive and Baidu Netdisk (Access code: fvxc).
Some examples from the TI-FRD dataset, showing campus scenes with diverse layouts and structures.
Comparison of data acquisition viewpoints in TI-FRD and ViViD++. TI-FRD enables flexible, multi-directional observations, whereas ViViD++'s driving sequences are captured by vehicle-mounted cameras with fixed perspectives. Arrows denote the viewing direction in each image.
The figure shows example RGB images, thermal images, and the corresponding ground-truth depth maps captured under both daytime and nighttime conditions.
Sensor specification for our TI-FRD dataset. acc. and accel. mean range accuracy and acceleration.
