Under Construction..
Mission:
News:
New dataset is captured by new multi-sensor system.
OUR LATEST RESEARCH
This paper introduces all-day dataset captured from KAIST campus for use in mobile robotics, autonomous driving, and recognition researches. Totally, we captured 42 km sequences at 15~100Hz using multiple sensor modalities such as fully aligned visible and thermal devices, high resolution stereo visible cameras, and a high accuracy GPS/IMU inertial navigation system. Despites of a particular scenario, we provide the first aligned visible/thermal all-day dataset, including various illumination conditions: day, night, sunset, and sunrise. With this dataset, we introduce multi-spectral loop-detector as a baseline. We will open all calibrated and synchronized datasets1, and hope to make a various state of the art computer vision and robotics algorithms.
As increasing of interest in pedestrian detection, the dataset has also been subject to the research in the past decades. However, most of existing datasets focus on color channel whereas the thermal channel is helpful clue for the detection even in the dark environment. In these respects, we propose a multispectral pedestrian dataset which provides well aligned color-thermal image pairs captured by a beam splitter based special hardware. The dataset is as large as previous color-based datasets and provides dense annotations including temporal correspondences. With this dataset, we introduce multispectral ACF which is extension of aggregated channel features (ACF) to simultaneously handle color-thermal image pairs. Multispectral ACF reduces averaging miss rate of ACF by 15%, which achieves another breakthrough in the pedestrian detection task.
