Berkeley MHAD: A comprehensive Multimodal Human Action Database

By Ferda Ofli, Rizwan Ahmed Chaudhry, Gregorij Kurillo, René Vidal, Ruzena Bajcsy. 2013

515 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Multimodal Learning, Evaluation & Benchmarks, Human-Computer Interaction

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 Multimodal Learning

In Evaluation & Benchmarks

In Human-Computer Interaction

Citations in the corpus

Co-Occurrence Feature Learning for Skeleton Based Action Recognition Using Regularized Deep LSTM Networks (2016, methodology): “We have also tested our model on the Berkeley MHAD action recognition dataset (Oﬂi et al. 2013) and achieved 100% accuracy.”
RGB-D-based action recognition datasets: A survey (2016, background): “Berkeley MHAD Berkeley Multimodal Human Action Database (MHAD) [42]”
Space-Time Representation of People Based on 3D Skeletal Data: A Review (2016, background): “Many other datasets captured using a Kinect v1 camera were also released to the public, including the MSR Daily Activity 3D [130], MSR Action Pairs [122], Online RGBD Action (ORGBD) [116], UTKinect-Action [133], Florence 3D-Action [127], CMU-MAD [113], UTD-MHAD [112], G3D/G3Di [128, 114], SPHERE [11”
RGB-D-based Human Motion Recognition with Deep Learning: A Survey (2017, background): “Most of available datasets require subjects to perform actions in a visible and restricted view to avoid occlusion, and this results in limited view data collection and less 29 SBU Kinect Interaction [149] Depth+ Skeleton PRNN Side Information 89.20% [130] Skeleton Co-occurrence +LSTM None 90.41% [1”
PKU-MMD: A Large Scale Benchmark for Continuous Multi-Modal Human Action Understanding (2017, neutral): “MSR-Action3D [17] 20 567 567 1 D+Skeleton No 2010 RGBD-HuDaAct [21] 13 1189 1189 1 RGB+D No 2011 MSR-DailyActivity [43] 16 320 320 1 RGB+D+Skeleton No 2012 Act4 [8] 14 6844 6844 1 RGB+D No 2012 UTKinect-Action [53] 10 200 200 1 RGB+D+Skeleton No 2012 3D Action Pairs [23] 12 360 360 1 RGB+D+Skeleton ”
Modular meta-learning (2018, methodology): “ses, the variance of the distribution decreases and modules become more specialized. (b) benchmarked robotic domains: MIT push dataset [28](top), action (throwing a ball) in the Berkeley MHAD dataset [29] (bottom). Given particular structures S j, then for each task T j, we know the parametric form ”
Deformable Shape Completion with Graph Convolutional Autoencoders (2017, methodology): “To this end, we make use of the MHAD dataset [39] which provides Kinect scans from 2 viewpoints of subjects performing a variety of actions.”
3D Flow Estimation for Human Action Recognition from Colored Point Clouds (2013, methodology): “Currently, the following datasets have been released: RGBD-HuDaAct Database (Zhang & Parker, 2011), Indoor Activity Database (Sung et al., 2012), MSR-Action3D Dataset (Wanqing Li, 2010), MSRDailyActivity3D Dataset (Jiang Wang & Yuan, 2012), LIRIS Human Activities Dataset (Wolf et al., 2012) and Berk”
Human Action Recognition Based on MOCAP Information Using Convolution Neural Networks (2014, methodology): “The proposed approach for human action recognition from MOCAP information using convolutional neural network is tested on Berkeley MHAD dataset [15].”
RGBD Datasets: Past, Present and Future (2016, methodology): “Four datasets stand out for capturing humans with a full MoCap setup [23, 34, 83, 51], while the Manipulation Action Dataset [1] is unique in providing semantic segmentation of objects as they are manipulated.”
QuaterNet: A Quaternion-based Recurrent Model for Human Motion (2018, unknown): “The training set comprises motion capture data from multiple sources [1, 39, 41, 62] at 120 Hz, and is re-targeted to a common skeleton.”
Hierarchical recurrent neural network for skeleton based action recognition (2015, methodology): “Berkeley MHAD: We follow the experimental protocol proposed in [22] on this dataset.”
Learning Clip Representations for Skeleton-Based 3D Action Recognition (2018, methodology): “6) Results on the Berkeley MHAD Dataset: For this dataset, we follow the standard testing protocol in [64] where the skeleton sequences of the first 7 subjects are used for training and the sequences of the other 5 subjects are used for testing.”
PiGraphs (2016, background): “, folding chairs [Ofli et al. 2013]) or specialized setups (e.”
Spatio-Temporal Naive-Bayes Nearest-Neighbor (ST-NBNN) for Skeleton-Based Action Recognition (2017, methodology): “Comparison results on the MSR-Action3D dataset [12], the UTKinect dataset [30], and the Berkeley MHAD dataset [17] are provided and discussed in Sec.”