Hyunggic !



Pedestrian DPM


Introduction
To train a pedestrian model, we used the Caltech Pedestrian Dataset which offers an opportunity to exploit many different aspects of model training thanks to its large number of positive samples. The (annotated) dataset corresponds to approximately 2.5 hours of video captured at 30fps. The dataset is segmented into 11 sessions, 6 of which were used for training (S0~S5) and the rest sessions were used for testing (S6~S10). For details, please refer to:

http://www.vision.caltech.edu/Image_Datasets/CaltechPedestrians/



[Trained Models with Different Part Number]


[Detection Result using the Caltech Benchmark]