MoSIFT: Recognizing Human Actions in Surveillance Videos

By Ming-yu Chen, Alexander G. Hauptmann. 2009

403 citations recorded by the source corpus.

How researchers see this paper

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

1. Yi Yang
2. Chuang Gan
3. Dacheng Tao
4. Wanli Ouyang
5. Luc Van Gool
6. X. Wang
7. Shuicheng Yan
8. Abhinav Kumar Gupta

Citations in the corpus

• Towards wearable cognitive assistance (2014, methodology): “Motion Classifier: To interpret motion in the surroundings (such as someone is waving to a user or running towards him), we have implemented a cognitive engine based on the MoSIFT algorithm [33].”
• Multi-Class Active Learning by Uncertainty Sampling with Diversity Maximization (2015, methodology): “In the experiment, we use the MoSift (Chen and Hauptmann 2009) descriptor, based on which a 32,768 dimension spatial BoW feature is computed to represent each video sequence (Yang et al.”
• An Adaptive Semisupervised Feature Analysis for Video Semantic Recognition (2018, methodology): “In the experiments, MoSIFT feature with 1,000-dimensionality is extracted to represent each video sequence for its consideration of both local appearance and local motion [45].”
• DevNet: A Deep Event Network for multimedia event detection and evidence recounting (2015, methodology): “Our framework relates to three research directions: event detection, event recounting, and CNNs, which will be briefly reviewed in this section.”
• Harnessing Lab Knowledge for Real-World Action Recognition (2014, background): “(1): min fl , fr nl∑ i=1 loss ( fl(x̃ i l ), y i l ) + nr∑ j=1 loss ( fr (x̃ j r ), y j r ) +α ( fl , fr ), (2) in which {x̃ il , yil }nli=1 and fl denote the labeled lab data and the correspondent classifier.”
• Odessa: enabling interactive perception applications on mobile devices (2011, background): “The latter accumulates frame pairs, and then extracts SIFT-like features that encode optical flow in addition to appearance [6].”
• A discriminative CNN video representation for event detection (2014, methodology): “MoSIFT [7] with Fisher vector achieves mAP 18.1% on 100Ex and 5.3% on 10Ex; STIP [22] with Fisher vector achieves mAP 15.0% on 100Ex and 7.1% on 10Ex; CSIFT [40] with Fisher vector achieves mAP 14.7% on 100Ex and 5.3% on 10Ex.”
• One-shot learning gesture recognition from RGB-D data using bag of features (2013, methodology): “Second, from the previous works, we know that traditional features have achieved promising results (Dollár et al., 2005; Laptev, 2005; Chen and Hauptmann, 2009).”
• Violence Detection in Video Using Computer Vision Techniques (2011, methodology): “MoSIFT [4] is an extension of the popular SIFT [17] image descriptor for video.”
• 3D SMoSIFT: three-dimensional sparse motion scale invariant feature transform for activity recognition from RGB-D videos (2014, methodology): “After keypoint detection, a feature descriptor can be computed from the local patch around a keypoint, such as Cuboid descriptor,(16) HOG and HOF descriptors,(17) or SIFT-like descriptors.(13,14,24) Here, we calculate SIFT-like Fig.”