TY - JOUR
T1 - Virtual and real world adaptationfor pedestrian detection
AU - Vazquez, David
AU - Lopez, Antonio M.
AU - Marin, Javier
AU - Ponsa, Daniel
AU - Geronimo, David
PY - 2014/1/1
Y1 - 2014/1/1
N2 - Pedestrian detection is of paramount interest for many applications. Most promising detectors rely on discriminatively learnt classifiers, i.e., trained with annotated samples. However, the annotation step is a human intensive and subjective task worth to be minimized. By using virtual worlds we can automatically obtain precise and rich annotations. Thus, we face the question: can a pedestrian appearance model learnt in realistic virtual worlds work successfully for pedestrian detection in real-world images? Conducted experiments show that virtual-world based training can provide excellent testing accuracy in real world, but it can also suffer the data set shift problem as real-world based training does. Accordingly, we have designed a domain adaptation framework, V-AYLA, in which we have tested different techniques to collect a few pedestrian samples from the target domain (real world) and combine them with the many examples of the source domain (virtual world) in order to train a domain adapted pedestrian classifier that will operate in the target domain. V-AYLA reports the same detection accuracy than when training with many human-provided pedestrian annotations and testing with real-world images of the same domain. To the best of our knowledge, this is the first work demonstrating adaptation of virtual and real worlds for developing an object detector. © 2014 IEEE.
AB - Pedestrian detection is of paramount interest for many applications. Most promising detectors rely on discriminatively learnt classifiers, i.e., trained with annotated samples. However, the annotation step is a human intensive and subjective task worth to be minimized. By using virtual worlds we can automatically obtain precise and rich annotations. Thus, we face the question: can a pedestrian appearance model learnt in realistic virtual worlds work successfully for pedestrian detection in real-world images? Conducted experiments show that virtual-world based training can provide excellent testing accuracy in real world, but it can also suffer the data set shift problem as real-world based training does. Accordingly, we have designed a domain adaptation framework, V-AYLA, in which we have tested different techniques to collect a few pedestrian samples from the target domain (real world) and combine them with the many examples of the source domain (virtual world) in order to train a domain adapted pedestrian classifier that will operate in the target domain. V-AYLA reports the same detection accuracy than when training with many human-provided pedestrian annotations and testing with real-world images of the same domain. To the best of our knowledge, this is the first work demonstrating adaptation of virtual and real worlds for developing an object detector. © 2014 IEEE.
KW - data set shift
KW - domain adaptation
KW - Pedestrian detection
KW - photo-realistic computer animation
U2 - 10.1109/TPAMI.2013.163
DO - 10.1109/TPAMI.2013.163
M3 - Article
SN - 0162-8828
VL - 36
SP - 797
EP - 809
JO - IEEE Transactions on Pattern Analysis and Machine Intelligence
JF - IEEE Transactions on Pattern Analysis and Machine Intelligence
IS - 4
M1 - 6587038
ER -