DevNet: A Deep Event Network for multimedia event detection and evidence recounting

By Chuang Gan, Naiyan Wang, Yi Yang, Dit-Yan Yeung, Alexander G. Hauptmann. 2015

336 citations recorded by the source corpus.

Subfields: Artificial Intelligence, 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 Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Deep Multimodal Feature Analysis for Action Recognition in RGB+D Videos (2016, background): “A novel deep convoultional framework for video event detection and evidence recounting was proposed by [12].”
Actions ~ Transformations (2015, methodology): “In this ex-periment, we visualize what the models focus on during inference using similar techniques as [38, 7].”
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].”
Revisiting the Effectiveness of Off-the-shelf Temporal Modeling Approaches for Large-scale Video Classification (2017, positive): “To address this challenge, our solution follows the strategy of DevNet framework [3].”
Towards Interpretable R-CNN by Unfolding Latent Structures (2017, background): “…in the high-scoring input image patches (Zeiler and Fergus 2014), identifying items in a visual scene and recount multimedia events (Yu et al. 2012; Gan et al. 2015), generating synthesized images by maximizing the response of a given node in the network (Erhan et al. 2009; Le et al. 2012;…”
A discriminative CNN video representation for event detection (2014, neutral): “crossframe max-pooling [11], fine-tuning could be helpful.”
Recognizing and Curating Photo Albums via Event-Specific Image Importance (2017, methodology): “The most popular approach uses videos as input [7, 17, 18, 22, 26, 29, 30], and spatiotemporal features are commonly used.”
Interpretable R-CNN (2017, methodology): “There are a variety of methods including identifying high-scoring image patches [23, 52] or over-segmented atomic regions [62] directly, visualizing the layers of convolutional networks using deconvolutional networks to understand what contents are emphasized in the high-scoring input image patches ”
On-Device Training Under 256KB Memory (2022, background): “The most straightforward way is to only update the classifier layer [16, 24, 27, 62], but the accuracy is low when the domain shift is large [13].”
Factors in Finetuning Deep Model for Object Detection with Long-Tail Distribution (2016, background): “eNet dataset is found to yield state-of-the-art performance for many vision tasks such as tracking [27], segmentation [16, 10], object detection [10, 24], action recognition [13], and event detection [8]. When ﬁnetuning the deep model for object detection, however, we have two observations. The ﬁrst”
Leveraging Structural Context Models and Ranking Score Fusion for Human Interaction Prediction (2016, background): “Pre-trained CNN models have been shown to be transferable across domains [39], and have outperformed hand-crafted features in a variety of visual recognition tasks [40]–[45].”
Appearance-and-Relation Networks for Video Classification (2017, background): “There are three kinds of successful architectures or frameworks for video classification [2]: (1) two-stream CNNs [35], (2) 3D CNNs [18, 41], and (3) 2D CNNs with temporal models on top such as LSTM [5, 33], temporal convolution [33], sparse sampling and aggregation [49], and attention modeling [48,”
VQS: Linking Segmentations to Questions and Answers for Supervised Attention in VQA and Question-Focused Semantic Segmentation (2017, background): “This is not surprising since latent variables hardly match semantic interpretations due to the lack of explicit training signals; similar observations exist in other studies, e.g., object detection [8], video recognition [11] and text processing [49].”
Leveraging Weak Semantic Relevance for Complex Video Event Classification (2017, neutral): “[5] fine-tune a CNN that are pre-trained”
Actionness Estimation Using Hybrid Fully Convolutional Networks (2016, background): “Convolutional neural networks (CNNs) have achieved remarkable successes for various tasks, such as object recognition [19, 31, 33, 49], event recognition [43, 8], crowd analysis [27, 28] and so on.”
Deep Learning for Computer Vision: A Brief Review (2018, methodology): “In [89] deep learning was used for complex event detection and recognition in video sequences: first, saliencymapswere used”
Improved Dense Trajectory with Cross Streams (2016, critical): “However, TDD, which is based on iDT and the two-stream approach, has two main shortcomings: (1) as shown in Figure 1, iDT cannot completely remove the background image for videos captured by a shaking camera.”
TURN TAP: Temporal Unit Regression Network for Temporal Action Proposals (2017, methodology): “rable work in action classiﬁcation task where a “trimmed” video is classiﬁed into one of speciﬁed categories [24,29]. There has also been work on localizing the actions in a longer, “untrimmed” video [6,23,35,33], i.e. temporal action localization. A straightforward way to use action classiﬁcation t”
Graph Convolutional Networks for Temporal Action Localization (2019, background): “Understanding human actions in videos has been becoming a prominent research topic in computer vision, owing to its various applications in security surveillance, human behavior analysis and many other areas [10, 35, 38, 12, 13, 14, 15, 16, 42].”