What do 15,000 object categories tell us about classifying and localizing actions?

By Mihir Jain, Jan van Gemert, Cees G. M. Snoek. 2015

190 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

Not Using the Car to See the Sidewalk — Quantifying and Controlling the Effects of Context in Classification and Segmentation (2018, background): “In computer vision, incorporating context information has been shown to improve performance in various tasks including object recognition [12, 22, 18] and action recognition [9], object detection [3] and segmentation [25].”
Chained Multi-stream Networks Exploiting Pose, Motion, and Appearance for Action Classification and Detection (2017, methodology): “In addition to optical flow, some works made use of other modalities like audio [50], warped flow [47], and object information [12] to capture complementary information for video classification.”
Action Recognition using Visual Attention (2015, background): “(Jain et al., 2015) 71.3 66.4 models in the case of HMDB-51 and Hollywood2.”
Temporal Convolutional Networks for Action Segmentation and Detection (2016, background): “Large-scale Recognition: There has been substantial work on spatiotemporal models for large scale video classification and detection [31, 11, 12, 27, 33, 23, 19].”
VideoLSTM convolves, attends and flows for action recognition (2016, methodology): “As a ﬁ nal experiment, we explore the bene ﬁ t of a late fusion of VideoLSTM action predictions with approaches relying on hand-crafted iDT features (Peng et al., 2014; Wang and Schmid, 2013) and object responses (Jain et al., 2015).”
Temporal Segment Networks for Action Recognition in Videos (2017, methodology): “DT+MVSV [46] 55:9% DT+MVSV [46] 83:5% object+motion [73] 71.”
Predicting the Where and What of Actors and Actions through Online Action Localization (2016, background): “Offline Action Localization has received significant attention in the past few years [41, 38, 9, 5, 13, 35, 11].”
Segmental Spatiotemporal CNNs for Fine-Grained Action Segmentation (2016, methodology): “Large-scale Action Classification: Recent efforts to extend CNN models to video [2,24,26,27,28,29,30] improve over holistic methods by encoding spatial and temporal information.”
Dynamic Sampling Networks for Efficient Action Recognition in Videos (2020, unknown): “1% Objects+Motion [59] RGB+HOG+HOF+MBH ALL 71.”
Leaving Some Stones Unturned: Dynamic Feature Prioritization for Activity Detection in Streaming Video (2016, methodology): “Object-Pref [4] is next best on ADL, though it is noticeably weaker on UCF because it does not account for the temporal redundancy of the dataset.”
Prompting Visual-Language Models for Efficient Video Understanding (2021, background): “One typical idea lies in learning a common representation space shared by seen and unseen categories, such as attributes space [20, 30, 48], semantic space [5, 21, 29, 42], synthesizing features [55], using objects to create common space [52].”
CDC: Convolutional-De-Convolutional Networks for Precise Temporal Action Localization in Untrimmed Videos (2017, methodology): “Various methods have been developed, from the perspective of supervoxel merging [20, 54, 55], tracking [69, 42, 62, 52], object detection and linking [28, 14, 75, 42, 62], spatio-temporal segmentation [31, 70], and leveraging still images [21, 58, 22].”
Efficient Action Detection in Untrimmed Videos via Multi-task Learning (2016, background): “Much progress has been made on action detection in manually trimmed, short videos over the past several years [47, 14, 13, 30, 39, 6] by employing handcrafted local features, such as improved dense trajectories (IDT) [35], or video representations that are learned directly from the data itself…”
Real-Time Action Recognition with Enhanced Motion Vector CNNs (2016, methodology): “Finally, we compare our speed and mean average precision with [8] on THUMOS14 dataset (Table 8).”