Shape-based pedestrian parsing

By Yihang Bo, Charless C. Fowlkes. 2011

86 citations recorded by the source corpus.

We describe a simple model for parsing pedestrians based on shape. Our model assembles candidate parts from an oversegmentation of the image and matches them to a library of exemplars. Our matching uses a hierarchical decomposition into a variable number of parts and computes scores on partial matchings in order to prune the search space of candidate segment. Simple constraints enforce consistent layout of parts. Because our model is shape-based, it generalizes well. We use exemplars from a controlled dataset of poses but achieve good test performance on unconstrained images of pedestrians in street scenes. We demonstrate results of parsing detections returned from a standard scanning-window pedestrian detector and use the resulting parse to perform viewpoint prediction and detection re-scoring.

Subfields: Computer Vision and Pattern Recognition, Ocean Engineering

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

Hypercolumns for object segmentation and fine-grained localization (2014, methodology): “There has also been work on parsing pedestrians [3, 31].”
Joint Object and Part Segmentation Using Deep Learned Potentials (2015, background): “However, due to its increased challenge in getting detailed boundaries, most previous work focused on parsing objects given both the semantic category and a cropped bounding box with no occlusion, such as human parsing [4, 45, 41, 13, 11], car parsing [35, 13, 28] or animal parsing [39].”
Zoom Better to See Clearer: Human and Object Parsing with Hierarchical Auto-Zoom Net (2015, background): “The study of human part parsing has been largely restricted to constrained environments, where a human instance in an image is well localized and has a relatively simple pose like standing or walking [3,36,9,34,8,21,33].”
Semantic part segmentation using compositional model combining shape and appearance (2014, background): “But there are only a few pieces of works on semantic part segmentation, such as human parsing [4, 10, 9, 29] and car parsing [26, 11, 23].”
Beyond Bounding Boxes: Precise Localization of Objects in Images (2015, background): “This is different from most prior work on these problems [57, 56, 60, 59, 4, 45], which operates on the immediate vicinity of ground-truth”
The Shape Boltzmann Machine: A Strong Model of Object Shape (2012, background): “…without internal structure, there are situations where it is desirable to explicitly model multiple, dependent regions, e.g. in order to decompose the foreground object into parts (Winn and Jojic, 2005; Kapoor and Winn, 2006; Thomas et al, 2009; Bo and Fowlkes, 2011; Eslami and Williams, 2011).”
Pedestrian Parsing via Deep Decompositional Network (2013, methodology): “We cast it as a data reconstruction problem using a strategy similar to DAE [26], which initializes the parameters by a RBM [8] with stochastically corrupted data as input, and then fine-tunes them by minimizing the square error between the reconstructed data and the clean data.”
Zoom Better to See Clearer: Human Part Segmentation with Auto Zoom Net (2015, background): “…into its semantic parts) is mostly addressed under constrained conditions such as presumably accurate localization, clear appearances, or relatively simple poses, e.g., (Bo & Fowlkes, 2011; Zhu et al., 2011; Eslami & Williams, 2012; Yamaguchi et al., 2012; Dong et al., 2014; Liu et al., 2015a).”
Iterated Second-Order Label Sensitive Pooling for 3D Human Pose Estimation (2014, methodology): “Body labeling methods have also been proposed in the context of person detection[4], to learn a model of the body[13] and recently for 2D pose prediction[21, 11], but not within models and algorithms carrying all the way to 3D pose estimation.”
Deep Learning for Semantic Part Segmentation with High-Level Guidance (2015, background): “In Bo & Fowlkes (2011), the authors compute appearance and shape features on oversegmentations of cropped pedestrian images from the Penn-Fudan pedestrian dataset Wang et al. (2007); Bo & Fowlkes (2011).”
Clothing Co-parsing by Joint Image Segmentation and Labeling (2014, background): “Recent approaches incorporated shape-based human model [2], or pose estimation and supervised region labeling [30], and achieved impressive results.”