#temporal modeling
7 resultsEnhanced 3D convolutional networks for crowd counting
Zhikang Zou, Huiliang Shao, Xiaoye Qu +2
The paper introduces a temporal channel-aware (TCA) block that uses 3D convolutions and channel-wise modulation to capture spatio‑temporal information for more accurate crowd count…
Audio-Visual Embedding for Cross-Modal MusicVideo Retrieval through Supervised Deep CCA
Donghuo Zeng, Yi Yu, Keizo Oyama
The paper introduces a supervised deep CCA model that embeds audio and video into a shared space for cross‑modal music video retrieval, using an attention‑based LSTM to select repr…
Tracking Temporal Evolution of Network Activity for Botnet Detection
Kapil Sinha, Arun Viswanathan, Julian Bunn
The paper introduces a supervised LSTM-based model that tracks the temporal evolution of network activity to detect botnet hosts, achieving 96.2% accuracy on the CTU-13 dataset.
Joint Group Feature Selection and Discriminative Filter Learning for Robust Visual Object Tracking
Tianyang Xu, Zhen-Hua Feng, Xiao-Jun Wu +1
The paper introduces a Group Feature Selection method for Discriminative Correlation Filters that jointly selects channel and spatial features and learns filters with temporal smoo…
Multi-Agent Reinforcement Learning Based Frame Sampling for Effective Untrimmed Video Recognition
Wenhao Wu, Dongliang He, Xiao Tan +2
The paper proposes a learning-based frame sampling method for untrimmed video recognition using multi-agent reinforcement learning to select informative frames, improving accuracy…
Temporal Gaussian Mixture Layer for Videos
AJ Piergiovanni, Michael S. Ryoo
The paper introduces the Temporal Gaussian Mixture (TGM) layer, a compact temporal convolutional layer that uses learnable Gaussian parameters to capture long-range temporal inform…
On the difficulty of learning and predicting the long-term dynamics of bouncing objects
Alberto Cenzato, Alberto Testolin, Marco Zorzi
The paper evaluates several unsupervised deep learning models on synthetic videos of bouncing objects, finding that while they predict the next frame well, they struggle to generat…