Wei Ke

Pedestrian Detection with Deep Convolutional Neural Network

The problem of pedestrian detection in image and video frames has been extensively investigated in the past decade. However, the low performance in complex scenes shows that it remains an open problem. In this paper, we propose to cascade simple Aggregated Channel Features (ACF) and rich Deep Convolutional Neural Network (DCNN) features for efficient and effective pedestrian detection in complex scenes. The ACF based detector is used to generate candidate pedestrian windows and the rich DCNN features are used for fine classification. Experiments show that the proposed approach achieved leading performance in the INRIA dataset and comparable performance to the state-of-the-art in the Caltech and ETH datasets.

SRN: Side-Output Residual Network for Object Symmetry Detection in the Wild

In this paper, we establish a baseline for object symmetry detection in complex backgrounds by presenting a new benchmark and an end-to-end deep learning approach, opening up a promising direction for symmetry detection in the wild. The new benchmark, named Sym-PASCAL, spans challenges including object diversity, multi-objects, part-invisibility, and various complex backgrounds that are far beyond those in existing datasets. The proposed symmetry detection approach, named Side-output Residual Network (SRN), leverages output Residual Units (RUs) to fit the errors between the object symmetry ground-truth and the outputs of RUs. By stacking RUs in a deep-to-shallow manner, SRN exploits the flow of errors among multiple scales to ease the problems of fitting complex outputs with limited layers, suppressing the complex backgrounds, and effectively matching object symmetry of different scales. Experimental results validate both the benchmark and its challenging aspects related to real-world images, and the state-of-the-art performance of our symmetry detection approach. The benchmark and the code for SRN are publicly available at https://github.com/KevinKecc/SRN.

A cross structured light sensor for weld line detection on wall-climbing robot

In this paper, we design a cross structured light (CSL) sensor for weld line detection on wall-climbing robot platform. The detection and measurement results can navigate robot to move along weld lines. CSL sensor consists of a CSL projector and a camera. The captured cross laser stripe represents the convexities of horizontal and vertical weld lines simultaneously. The sensor is calibrated by a planar target. The experiment results show that the sensor can accurately obtain 3D information of weld line with very low measurement errors.

Pedestrian detection via PCA filters based convolutional channel features

In this paper, we propose a kind of image representation, named PCA filters based convolutional channel features (PCA-CCF) for pedestrian detection. The motivation is to use the convolutional network architecture with orthogonal PCA filters to enhance the state-of-the-art aggregate channel features (ACF). In PCA-CCF, the convolutional operation improves the feature robustness to pedestrian local deformation. The learned PCA filters reduce the correlations among features of each channel, and therefore, improve feature discrimination capability. With the proposed PCA-CCF features and cascaded AdaBoost classifiers, we develop a coarse-to-fine pedestrian detection approach. Experiments show that such approach achieves 3.04%, 17.87% and 6.28% performance gain on the INRIA, Caltech Reasonable and Caltech Overall pedestrian datasets, respectively.

Self-Learning Scene-Specific Pedestrian Detectors Using a Progressive Latent Model

In this paper, a self-learning approach is proposed towards solving scene-specific pedestrian detection problem without any human annotation involved. The self-learning approach is deployed as progressive steps of object discovery, object enforcement, and label propagation. In the learning procedure, object locations in each frame are treated as latent variables that are solved with a progressive latent model (PLM). Compared with conventional latent models, the proposed PLM incorporates a spatial regularization term to reduce ambiguities in object proposals and to enforce object localization, and also a graph-based label propagation to discover harder instances in adjacent frames. With the difference of convex (DC) objective functions, PLM can be efficiently optimized with a concave-convex programming and thus guaranteeing the stability of self-learning. Extensive experiments demonstrate that even without annotation the proposed self-learning approach outperforms weakly supervised learning approaches, while achieving comparable performance with transfer learning and fully supervised approaches.

Minimum Entropy Models for Laser Line Extraction

Minimum entropy model can find the optimal gray space for laser line extraction. A global model named Minimum Entropy Deconvolution is established to search for the peaks which constitute the laser line. Not only does it reach a high accuracy, but also it retains the line smoothness, which the previous work often paid little attention to. Besides, this work could extract several laser lines. Experimental results show that the robust models and fast algorithms outperform the compared.

Texture Complexity Based Redundant Regions Ranking for Object Proposal

Object proposal has been successfully applied in recent visual object detection approaches and shown improved computational efficiency. The purpose of object proposal is to use as few as regions to cover as many as objects. In this paper, we propose a strategy named Texture Complexity based Redundant Regions Ranking (TCR) for object proposal. Our approach first produces rich but redundant regions using a color segmentation approach, i.e. Selective Search. It then uses Texture Complexity (TC) based on complete contour number and Local Binary Pattern (LBP) entropy to measure the objectness score of each region. By ranking based on the TC, it is expected that as many as true object regions are preserved, while the number of the regions is significantly reduced. Experimental results on the PASCAL VOC 2007 dataset show that the proposed TCR significantly improves the baseline approach by increasing AUC (area under recall curve) from 0.39 to 0.48. It also outperforms the state-of-the-art with AUC and uses fewer detection proposals to achieve comparable recall rates.