Pedestrian Alignment Network for Large-scale Person Re-Identification

By Zhedong Zheng, Liang Zheng, Yi Yang. IEEE transactions on circuits and systems for video technology (Print), 2017

492 citations recorded by the source corpus.

Person re-identification (re-ID) is mostly viewed as an image retrieval problem. This task aims to search a query person in a large image pool. In practice, person re-ID usually adopts automatic detectors to obtain cropped pedestrian images. However, this process suffers from two types of detector errors: excessive background and part missing. Both errors deteriorate the quality of pedestrian alignment and may compromise pedestrian matching due to the position and scale variances. To address the misalignment problem, we propose that alignment be learned from an identification procedure. We introduce the pedestrian alignment network (PAN) which allows discriminative embedding learning pedestrian alignment without extra annotations. We observe that when the convolutional neural network learns to discriminate between different identities, the learned feature maps usually exhibit strong activations on the human body rather than the background. The proposed network thus takes advantage of this attention mechanism to adaptively locate and align pedestrians within a bounding box. Visual examples show that pedestrians are better aligned with PAN. Experiments on three large-scale re-ID datasets confirm that PAN improves the discriminative ability of the feature embeddings and yields competitive accuracy with the state-of-the-art methods.

Subfields: Computer Vision and Pattern Recognition, Ocean Engineering

View Paper: https://www.semanticscholar.org/paper/b856c0eb039effce7da9ff45c3f5987f18928bef

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Vision and Pattern Recognition

In Ocean Engineering

Citations in the corpus

Beyond Part Models: Person Retrieval with Refined Part Pooling (2017, methodology): “We note that the IDE model specified in [40] is a commonly used baseline in deep re-ID systems [40, 38, 33, 10, 28, 42, 43, 45].”
Each Part Matters: Local Patterns Facilitate Cross-View Geo-Localization (2020, methodology): “Different from works of the second line [38], [41], [42], [43], our partition method makes the network have good scalability to image rotation (see Figure 2).”
In Defense of the Triplet Loss for Person Re-Identification (2017, methodology): “We do not include approaches which are orthogonal to ours and could be integrated in a straightforward manner, such as various re-ranking schemes, data augmentation, and regularization [3, 43, 54, 7, 47, 53, 56].”
Pyramidal Person Re-IDentification via Multi-Loss Dynamic Training (2018, background): “A pedestrian alignment network [37] is built to learn discriminative embedding and pedestrian alignment without extra annotations.”
Attention mechanisms in computer vision: A survey (2021, unknown): “[97, 46, 139, 70, 128, 36, 11, 156] , [137, 143, 63, 161, 159, 6, 177, 80] , [67, 180, 10, 23, 24, 163, 155, 151] , [44, 49, 106, 64, 168, 18, 11, 36] , [48, 16, 169, 153, 174, 127, 53, 92] , [89, 21, 5, 134, 65, 171, 2, 109] , [45, 82, 157, 91, 100, 129, 146, 131] , [138, 85, 175, 76, 166, 133, 167”
Horizontal Pyramid Matching for Person Re-identification (2018, methodology): “For the identiﬁcation model (Xiao et al. 2016; Zheng, Zheng, and Yang 2017a; Sun et al. 2017b), it tries to learn a discriminative representation of given input image and it always yields superior accuracy compared with veriﬁcation model.”
Features for Multi-target Multi-camera Tracking and Re-identification (2018, methodology): “This idea has been shown to work for person detection [17], and implicitly pursued in single-camera trackers [66, 67, 73] and Re-ID methods [35, 82, 83] that improve features to increase accuracy.”
FD-GAN: Pose-guided Feature Distilling GAN for Robust Person Re-identification (2018, methodology): “PAN [1] aligns pedestrians and learn person features simultaneously without any extra annotation.”
Attention-Aware Compositional Network for Person Re-identification (2018, methodology): “One set is poseirrelevant, which includes the null space semi-supervised learning method (NFST) [49], the domain guided dropout method (JSTL) [44], the deep transfer learning method (Transfer) [16], the Singular Vector Decomposition method (SVDNet) [39], the Online Instance Matching (OIM) method [45”
Pose-Guided Feature Alignment for Occluded Person Re-Identification (2019, unknown): “We [42] 81.0 63.4 - - Pedestrian[47] 82.0 63.0 - - DSR [11] 83.5 64.2 - - MultiLoss [35] 83.9 64.4 - - TripletLoss [13] 84.9 69.1 - - Adver Occluded [14] 86.5 78.3 79.1 62.1 APR [22] 87.0 66.9 73.9 55.6 MultiScale [4] 88.9 [32].”