Recognizing human actions: a local SVM approach

By Christian Schüldt, Ivan Laptev, Barbara Caputo. 2004

4,208 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition

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

Citations in the corpus

UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild (2012, background): “…action recognition datasets suffer from two disadvantages: 1) The number of their classes is typically very low compared to the richness of performed actions by humans in reality, e.g. KTH [11], Weizmann [3], UCF Sports [10], IXMAS [12] datasets includes only 6, 9, 9, 11 classes respectively.”
Dense-Captioning Events in Videos (2017, background): “he workers who completed our tasks. 7.1. Comparison to other datasets. Curation and open distribution is closely correlated with progress in the ﬁeld of video understanding (Table 4). The KTH dataset [42] pioneered the ﬁeld by studying human actions with a black background. Since then, datasets like”
HMDB: A large video database for human motion recognition (2011, background): “State-of-the-art performance on these datasets is now near ceiling and thus there is a need for the design and creation of new benchmarks.”
Hollywood in Homes: Crowdsourcing Data Collection for Activity Understanding (2016, background): “The KTH action dataset [12] paved the way for algorithms that recognized human actions.”
AVA: A Video Dataset of Spatio-Temporally Localized Atomic Visual Actions (2017, background): “Most popular action clas-siﬁcation datasets, such as KTH [35], Weizmann [4], Hollywood-2 [26], HMDB [24], UCF101 [39] consist of short clips, manually trimmed to capture a single action.”
Continual Lifelong Learning with Neural Networks: A Review (2018, neutral): “inimizing interference by creating new neurons whenever they are required. This approach has shown competitive results with batch learning methods on the Weizmann (Gorelick et al., 2005) and the KTH (Schuldt et al., 2004) action benchmark datasets. Furthermore, it learns robust action-label mappings”
Lifelong learning of human actions with deep neural network self-organization (2017, methodology): “We compare the classification accuracy of our model to stateof-the-art learning methods on the Weizmann (Gorelick et al., 2005) and the KTH (Schuldt et al., 2004) action benchmarks, showing competitive performance for the batch learning scenario with different evaluation schemes.”
Differential Recurrent Neural Networks for Action Recognition (2015, methodology): “We choose KTH dataset [27] for it is a de facto benchmark for evaluating action recognition algorithms.”
C3D: Generic Features for Video Analysis (2014, background): “The computer vision community has been working on video analysis for decades and tackled different problems such as action recognition [2, 31, 40], abnormal event detection [3, 7, 55], and activity understanding [29].”
Long-Term Temporal Convolutions for Action Recognition (2016, neutral): “Such a view ﬁnds support both in psychology [1] and in computer vision approaches to action recognition in video [2], [3], [4], [5].”
Finding action tubes (2014, background): “[34] build local space-time features to recognize action patters using SVM classifiers.”
Video In Sentences Out (2012, background): “…McKevitt, 1994, 1995–1996) and recognition of action in video (Blank et al., 2005; Laptev et al., 2008; Liu et al., 2009; Rodriguez et al., 2008; Schuldt et al., 2004; Siskind and Morris, 1996; Starner et al., 1998; Wang and Mori, 2009; Xu et al., 2002, 2005) have been of considerable interest…”
Every Moment Counts: Dense Detailed Labeling of Actions in Complex Videos (2015, neutral): “The KTH dataset [24] was a catalyst for a body of work.”
The “Something Something” Video Database for Learning and Evaluating Visual Common Sense (2017, background): “Not surprisingly, most of the currently popular labeled video datasets are action recognition datasets [26, 17, 23, 12].”
Picture: A probabilistic programming language for scene perception (2015, methodology): “We evaluated our program on a dataset of humans performing a variety of poses, which was aggregated from KTH [39] and LabelMe [38] images with signiﬁcant occlusion in the “person sitting”(around 50 total images).”
Video Action Transformer Network (2018, background): “Datasets have become progressively larger and harder: from actors performing simple actions [12, 33], to short sports and movie clips [24, 38], ﬁnally to diverse youtube videos [1, 23].”
Moments in Time Dataset: One Million Videos for Event Understanding (2018, background): “The KTH [40] and Weizmann [8] were early datasets for human action understanding.”
Human Action Recognition Using Factorized Spatio-Temporal Convolutional Networks (2015, background): “Possibly due to the increased complexity and difficulty of training 3D kernels without sufficient training video data (as compared to massive image datasets [3]), 3D CNN [7] does not perform well even on the less challenging KTH dataset [25].”