Searching Video for Complex Activities with Finite State Models

By Nazli Ikizler, D.A. Forsyth. 2007

112 citations recorded by the source corpus.

Subfields: Computer Vision and Pattern Recognition, Control and Systems Engineering

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 Control and Systems Engineering

Citations in the corpus

One-shot-learning gesture recognition using HOG-HOF features (2013, methodology): “Ikizler and Forsyth (2007) use the output of Human Motion Capture systems in combination with Hidden Markov Models.”
Human Activity Recognition with Metric Learning (2008, background): “At higher spatial resolutions one can recover body configuration and reason about it [2,3].”
3D skeleton-based human action classification: A survey (2016, background): “Therefore, there is a vast literature, including interesting surveys [13 – 19], on human body pose estimation [20 – 24] and activity/action classi ﬁ cation [25 – 30].”
Discovering discriminative action parts from mid-level video representations (2012, methodology): “Sequential data models have been employed to represent the temporal variability [11, 21].”
Human Action Recognition by Semilatent Topic Models (2009, unknown): “Ikizler and Forsyth [34] use finite state models to search for complex activities from videos.”
Making Action Recognition Robust to Occlusions and Viewpoint Changes (2010, methodology): “To handle occlusions, an alternative to global models is to use part based ones to make independent decisions for individual body parts and to fuse them into a global interpretation [7].”
A latent model of discriminative aspect (2009, background): “Second, one might build generative models from image data (for example, [8]), or motion capture data (for example, [16]).”
Understanding Video Events: A Survey of Methods for Automatic Interpretation of Semantic Occurrences in Video (2009, background): “Trajectories are another very popular object-based abstraction approach prevalent in the literature [45, 46, 47].”
Cross-dataset action detection (2010, background): “Models with STIPs avoid the alignment problem in temporal domain [5], and can effectively distinguish periodic or short-interval actions such as running and jogging with clean background.”
Learning to Recognize Activities from the Wrong View Point (2008, background): “Alternatively, Bregler fits a switching linear dynamical system ([10]; see also [37]); Ikizler and Forsyth build primitives explicitly by clustering motion capture [21].”
Inverse Dynamics for Action Recognition (2013, methodology): “Using both motion capture and video data, Ikizler and Forsyth [10] proposed a generative model to query complex activities in a large collection of videos.”
Adaptive Focus for Efficient Video Recognition (2021, background): “, on YouTube or TikTok) has fueled the demands for automatically recognizing human actions, events, or other contents within them, which benefits applications like recommendation [7, 8, 15], surveillance [6, 3] and content-based searching [22].”
Poselet Key-Framing: A Model for Human Activity Recognition (2013, methodology): “Sequential models: Hidden semi-Markov Models (HSMM) [10] , CRFs [31], and finite-state-machines [13] have been used to model the temporal evolution of human activities.”