Part-Based Feature Aggregation Method for Dynamic Scene Recognition

Existing methods for dynamic scene recognition mostly use global features extracted from the entire video frame or a video segment. In this paper, a part-based method is proposed for aggregating local features from multiple video frames. A pre-trained Fast R-CNN model is used to extract local convolutional layer features from the regions of interest (ROIs) of training images. These features are then clustered to locate representative parts. A set cover problem is formulated to select the discriminative parts, which are further refined by fine-tuning the Fast R-CNN. Local convolutional layer features and fully-connected layer features are extracted using the fine-tuned Fast R-CNN model, and then aggregated separately from a video segment to form two feature representations. They are concatenated into a global feature representation. Experimental results show that the proposed method outperforms several state-of-the-art features on two dynamic scene datasets.

• Publication year

  2019

• Venue

  International Conference on Digital Image Computing: Techniques and Applications

• Publication date

  2019-12-01

• Fields of study

  Computer Science

• Identifiers
  DOI 10.1109/DICTA47822.2019.8946036
• External record

  Open on Semantic Scholar

• Source metadata

  Semantic Scholar

• No claims are published for this paper.

• No concepts are published for this paper.

• Long-Short-Term Features for Dynamic Scene Classification

  2019influential reference
• A Super Descriptor Tensor Decomposition for Dynamic Scene Recognition

  2019cited by this paper
• Human Brain Tissue Segmentation in fMRI using Deep Long-Term Recurrent Convolutional Network

  2018cited by this paper
• D3: Recognizing dynamic scenes with deep dual descriptor based on key frames and key segments

  2017influential reference
• A Closer Look at Spatiotemporal Convolutions for Action Recognition

  2017influential reference
• Quo Vadis, Action Recognition? A New Model and the Kinetics Dataset

  2017influential reference
• Temporal Residual Networks for Dynamic Scene Recognition

  2017cited by this paper
• Dynamic Scene Recognition with Complementary Spatiotemporal Features

  2016cited by this paper
• Dynamic Scene Classification Using Redundant Spatial Scenelets

  2016cited by this paper
• SPLeaP: Soft Pooling of Learned Parts for Image Classification

  2016influential reference
• Learning Dictionary of Discriminative Part Detectors for Image Categorization and Cosegmentation

  2016cited by this paper
• Dynamic texture and scene classification by transferring deep image features

  2015cited by this paper
• Deep Residual Learning for Image Recognition

  2015influential reference
• Fast R-CNN

  2015cited by this paper
• Mid-level deep pattern mining

  2014cited by this paper
• CNN Features Off-the-Shelf: An Astounding Baseline for Recognition

  2014cited by this paper
• Learning Deep Features for Scene Recognition using Places Database

  2014influential reference
• Bags of Spacetime Energies for Dynamic Scene Recognition

  2014cited by this paper
• Learning Spatiotemporal Features with 3D Convolutional Networks

  2014influential reference
• Two-Stream Convolutional Networks for Action Recognition in Videos

  2014cited by this paper
• Very Deep Convolutional Networks for Large-Scale Image Recognition

  2014cited by this paper
• Learning Discriminative and Shareable Features for Scene Classification

  2014cited by this paper
• Large-Scale Video Classification with Convolutional Neural Networks

  2014influential reference
• Blocks That Shout: Distinctive Parts for Scene Classification

  2013cited by this paper
• Representing Videos Using Mid-level Discriminative Patches

  2013cited by this paper
• Selective Search for Object Recognition

  2013cited by this paper
• Dynamic Scene Classification: Learning Motion Descriptors with Slow Features Analysis

  2013cited by this paper
• Aggregating Local Image Descriptors into Compact Codes

  2012cited by this paper
• UCF101: A Dataset of 101 Human Actions Classes From Videos in The Wild

  2012cited by this paper
• Dynamic scene understanding: The role of orientation features in space and time in scene classification

  2012cited by this paper
• Spacetime Texture Representation and Recognition Based on a Spatiotemporal Orientation Analysis

  2012cited by this paper
• Unsupervised Discovery of Mid-Level Discriminative Patches

  2012cited by this paper
• Dynamic texture recognition based on distributions of spacetime oriented structure

  2010cited by this paper
• Object Detection with Discriminatively Trained Part Based Models

  2010cited by this paper
• Moving vistas: Exploiting motion for describing scenes

  2010influential reference
• Improving the Fisher Kernel for Large-Scale Image Classification

  2010cited by this paper
• ImageNet: A large-scale hierarchical image database

  2009influential reference
• Poselets: Body part detectors trained using 3D human pose annotations

  2009cited by this paper
• Recognizing indoor scenes

  2009cited by this paper
• LIBLINEAR: A Library for Large Linear Classification

  2008cited by this paper
• Author manuscript, published in "International Conference on Computer Vision (2013)" Action Recognition with Improved Trajectories

  year unknowncited by this paper

• Advancing Complex Wide-Area Scene Understanding with Hierarchical Coresets Selection

  2025cites this paper
• Feature Fusion of Deep and Spatio-Temporal Features for Dynamic Scene Understanding

  2025influential citation
• SAGN: Semantic-Aware Graph Network for Remote Sensing Scene Classification

  2023cites this paper
• A Novel Multi-Modal Network-Based Dynamic Scene Understanding

  2022cites this paper
• Cam-Net: Compressed Attentive Multi-Granularity Network For Dynamic Scene Classification

  2020cites this paper