Logo des Repositoriums

Auflistung nach Autor:in "Yaghoubi, Ehsan"

1 - 2 von 2
  • Konferenzbeitrag
    Pose Switch-based Convolutional Neural Network for Clothing Analysis in Visual Surveillance Environment
    (BIOSIG 2019 - Proceedings of the 18th International Conference of the Biometrics Special Interest Group, 2019) Alirezazadeh, Pendar; Yaghoubi, Ehsan; Assunção, Eduardo; Neves, João C.; Proença, Hugo
    Recognizing pedestrian clothing types and styles in outdoor scenes and totally uncontrolled conditions is appealing to emerging applications such as security, intelligent customer profile analysis and computer-aided fashion design. Recognition of clothing categories from videos remains a challenge, mainly due to the poor data resolution and the data covariates that compromise the effectiveness of automated image analysis techniques (e.g., poses, shadows and partial occlusions). While state-of-the-art methods typically analyze clothing attributes without paying attention to variation of human poses, here we claim for the importance of a feature representation derived from human poses to improve classification rate. Estimating the pose of pedestrians is important to fed guided features into recognizing system. In this paper, we introduce pose switch-based convolutional neural network for recognizing the types of clothes of pedestrians, using data acquired in crowded urban environments. In particular, we compare the effectiveness attained when using CNNs without respect to human poses variant, and assess the improvements in performance attained by pose feature extraction. The observed results enable us to conclude that pose information can improve the performance of clothing recognition system. We focus on the key role of pose information in pedestrian clothing analysis, which can be employed as an interesting topic for further works.
  • Konferenzbeitrag
    Region-Based CNNs for Pedestrian Gender Recognition in Visual Surveillance Environments
    (BIOSIG 2019 - Proceedings of the 18th International Conference of the Biometrics Special Interest Group, 2019) Yaghoubi, Ehsan; Alirezazadeh, Pendar; Assunção, Eduardo; Neves, João C.; Proença, Hugo
    Inferring soft biometric labels in totally uncontrolled outdoor environments, such as surveillance scenarios, remains a challenge due to the low resolution of data and its covariates that might seriously compromise performance (e.g., occlusions and subjects pose). In this kind of data, even state-of-the-art deep-learning frameworks (such as ResNet) working in a holistic way, attain relatively poor performance, which was the main motivation for the work described in this paper. In particular, having noticed the main effect of the subjects’ “pose” factor, in this paper we describe a method that uses the body keypoints to estimate the subjects pose and define a set of regions of interest (e.g., head, torso, and legs). This information is used to learn appropriate classification models, specialized in different poses/body parts, which contributes to solid improvements in performance. This conclusion is supported by the experiments we conducted in multiple real-world outdoor scenarios, using the data acquired from advertising panels placed in crowded urban environments.