UvA-DARE (Digital Academic Repository) Dynamic Image Networks for Action Recognition

By Hakan Bilen, Basura Fernando, Efstratios Gavves, Andrea Vedaldi, Stephen Gould. 2016

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

A Constructive Prediction of the Generalization Error Across Scales (2019, methodology): “, 2012): originally a video action recognition dataset, converted using the method of Bilen et al. (2016) into a single image per video.”
Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset (2017, background): “…for example by learning frame ranking functions for action classes and using these as a representation [9], by making analogies between actions and transformations [33], or by creating 2D visual snapshots of frame sequences [2] – this idea is related to the classic motion history work of [3].”
Rethinking Spatiotemporal Feature Learning: Speed-Accuracy Trade-offs in Video Classification (2017, background): “Since then, many video classiﬁcation methods follow the same multi-stream 2D CNN design, and have made improvements in terms of new representations [25,26], different backbone architecture [27,28,29,17], fusion of the streams [30,31,32,33] and exploiting richer temporal structures [34,35,36].”
UvA-DARE (Digital Academic Repository) Self-Supervised Video Representation Learning With Odd-One-Out Networks (positive): “Dynamic image [5, 4] encoder: This method is similar to the sum of differences of frames method, however the only difference is that now the input sequence is pre-processed to obtain a smoothed sequence M = 〈M1,M2, · · ·MW 〉.”
Spatiotemporal Residual Networks for Video Action Recognition (2016, methodology): “The combination of ConvNet methods with trajectory-based hand-crafted IDT features [29] typically boosts performance nontrivially [2, 26].”
Video Representation Learning Using Discriminative Pooling (2018, methodology): “For comparison, we plot the running times for some of the recent pooling schemes such as rank pooling [2, 14] and the Fisher vectors [51].”
RGB-D-based Human Motion Recognition with Deep Learning: A Survey (2017, methodology): “It has been shown that fine-tuning motion-based networks with spatial data (ImageNet) is more effective than training from scratch [56, 21, 51, 22].”
Scene Flow to Action Map: A New Representation for RGB-D Based Action Recognition with Convolutional Neural Networks (2017, methodology): “For SFAM-CTKRP, we revised the codes of paper [1] based on MatConvNet [40].”
Long-Term Temporal Convolutions for Action Recognition (2016, neutral): “Extensions of CNN representations to action recognition in video have been proposed in several recent works [6], [12], [13], [14], [15], [16], [22], [23], [24], [25].”
TSM: Temporal Shift Module for Efficient Video Understanding (2018, methodology): “Using the 2D CNN is a straightforward way to conduct video recognition [24, 39, 48, 11, 8, 9, 2].”
Moments in Time Dataset: One Million Videos for Event Understanding (2018, methodology): “ optical ﬂow). We use the BNInception [22] as the base model, by modifying the ﬁrst convolutional layer to accept 10 input channels instead of 3 as BNInception-Flow. Second, we compute dynamic images [7] as a means of spatiotemporal encoding of videos. A dynamic image summarizes the gist of a video ”
Learning and Using the Arrow of Time (2018, methodology): “In terms of the input modality, all three modalities have similar performance for action recognition using ResNet-50 with our AoT pre-training and also with Ima-geNet pre-training as shown in Bilen et al. [2].”
Adding New Tasks to a Single Network with Weight Trasformations using Binary Masks (2018, background): “Net [46], CIFAR-100 [22], Aircraft [31], Daimler pedestrian classication (DPed) [36], Describable textures (DTD) [5], German trac signs (GTSR) [51] , Omniglot [26], SVHN [38], UCF101 Dynamic Images [3,50] and VGG-Flowers [39]. A more detailed description of the challenge and the datasets can be foun”
Action Recognition Based on Joint Trajectory Maps with Convolutional Neural Networks (2016, background): “The last category is to represent a video in one or multiple compact images and adopt available trained ConvNet architectures for ﬁne-tuning [17], [51], [52], [77], [78].”
A Survey on Deep Learning Based Approaches for Action and Gesture Recognition in Image Sequences (2017, background): “[10] 2016 Dynamic feat maps BVLC CaffeNet 89.”
3D action recognition using data visualization and convolutional neural networks (2017, methodology): “[15] proposed a dynamic image representation, which is a single RGB image generated by applying approximate rank pooling operator on raw image pixels of a sequence.”