A Multilevel Mixture-of-Experts Framework for Pedestrian Classification

By Markus Enzweiler, Dariu M. Gavrila. 2011

170 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Model Architectures

How researchers see this paper

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

In Artificial Intelligence

In Computer Vision and Pattern Recognition

In Model Architectures

Citations in the corpus

VoxelNet: End-to-End Learning for Point Cloud Based 3D Object Detection (2017, neutral): “There are also several multi-modal fusion methods that combine images and LiDAR to improve detection accuracy [10, 16, 5].”
Ten Years of Pedestrian Detection, What Have We Learned? (2014, result): “Good success exploiting flow and stereo on the Daimler dataset has been reported [48], but similar results have yet to be seen on newer datasets such as KITTI.”
Making Bertha See (2013, methodology): “Furthermore, all features operate both on gray-level intensity as well as dense disparity images to fully exploit the orthogonal characteristics of both modalities [4], as shown in Figure 7.”
Making Bertha Drive—An Autonomous Journey on a Historic Route (2014, methodology): “Here, we are using a Mixture-of-Experts scheme that operates on a diverse set of image features and modalities as described in [18].”
Multiview random forest of local experts combining RGB and LIDAR data for pedestrian detection (2015, methodology): “in [30] multiples cues are combined following late-fusion style, while we consider also early-fusion, which, in fact, give better results in our framework.”
Computer Vision for Autonomous Vehicles: Problems, Datasets and State-of-the-Art (2017, background): “The combination of information from different sensors via sensor fusion (Enzweiler & Gavrila (2011); Chen et al. (2016b); González et al. (2016)) allows for the robust integration of this complementary information.”
Joint 3D Proposal Generation and Object Detection from View Aggregation (2017, background): “Multimodal object detectors have been previously discussed in the literature [12, 13, 14, 15, 3, 7], but up to our knowledge, no previous work has exploited features from both point cloud and RGB data to generate region proposals.”
Pedestrian Detection at Day/Night Time with Visible and FIR Cameras: A Comparison (2016, background): “Another recent trend has been to complement those appearance-based features, computed from single frames, with additional motion and depth features such as in [21–25].”
PointFusion: Deep Sensor Fusion for 3D Bounding Box Estimation (2017, methodology): “stage pipeline, which preprocesses each sensor modality separately and then performs a late fusion or decision-level fusion step using an expert-designed tracking system such as a Kalman filter [4, 7].”
Optimal Sensor Data Fusion Architecture for Object Detection in Adverse Weather Conditions (2018, methodology): “[4] detect pedestrians using a mixture-of-experts framework, in which RGB and depth images are fused in the early stage.”
On-Board Object Detection: Multicue, Multimodal, and Multiview Random Forest of Local Experts (2017, methodology): “types of features has been seen to boost the performance of either feature separately [11]–[13].”
Single-Pedestrian Detection Aided by Multi-pedestrian Detection (2013, background): “Features under investigation include Haar-like features [79], edgelets [87], shapelet [67], histogram of gradients (HOG) [10], dense SIFT [77], bag-of-words [43], integral histogram [66], color histogram [80], gradient histogram [99], covariance descriptor [76], co-occurrence features [69], local bi”
Efficient Multi-cue Scene Segmentation (2013, background): “The high density of the disparity maps allows us to extract complex features (such as SIFT) directly on the disparity image, similar to [6].”
Virtual and Real World Adaptation for Pedestrian Detection (2014, methodology): “We evaluate HOG and LBP separately instead of only considering the combination HOG+LBP, because we aim to assess the behavior of such single descriptors when transferred from virtual-world images to realworld ones; moreover they are used separately as experts by some mixture-of-experts pedestrian cl”