Dynamic Convolution: Attention Over Convolution Kernels

Chen, Yinpeng; Dai, Xiyang; Liu, Mengchen; Chen, Dongdong; Liu, Yuan; Liu, Zicheng

doi:10.1109/cvpr42600.2020.01104

Cited by 615 publications

(336 citation statements)

References 19 publications

Supporting

Mentioning

336

Contrasting

Order By: Relevance

“…To enable the network to fully learn effective information from the encoded features and upsampled features, we introduce dynamic convolution [38] modules in the decoding process to take advantage of these features adaptively. From the perspective of perception, the traditional or static perception used in other standard convolutional layers could be presented as y = g(W T x + b), where parameters W and b are weight matrix and bias, respectively.…”

Section: Dynamic Convolution Modulesmentioning

confidence: 99%

HDFNet: Hierarchical Dynamic Fusion Network for Change Detection in Optical Aerial Images

Zhang

et al. 2021

Remote Sensing

View full text Add to dashboard Cite

Accurate change detection in optical aerial images by using deep learning techniques has been attracting lots of research efforts in recent years. Correct change-detection results usually involve both global and local deep learning features. Existing deep learning approaches have achieved good performance on this task. However, under the scenarios of containing multiscale change areas within a bi-temporal image pair, existing methods still have shortcomings in adapting these change areas, such as false detection and limited completeness in detected areas. To deal with these problems, we design a hierarchical dynamic fusion network (HDFNet) to implement the optical aerial image-change detection task. Specifically, we propose a change-detection framework with hierarchical fusion strategy to provide sufficient information encouraging for change detection and introduce dynamic convolution modules to self-adaptively learn from this information. Also, we use a multilevel supervision strategy with multiscale loss functions to supervise the training process. Comprehensive experiments are conducted on two benchmark datasets, LEBEDEV and LEVIR-CD, to verify the effectiveness of the proposed method and the experimental results show that our model achieves state-of-the-art performance.

show abstract

Section: Dynamic Convolution Modulesmentioning

confidence: 99%

HDFNet: Hierarchical Dynamic Fusion Network for Change Detection in Optical Aerial Images

Zhang

et al. 2021

Remote Sensing

View full text Add to dashboard Cite

show abstract

“…Inspired by Squeeze-and-Excitation (SE) networks [51] and dynamic convolution [52], we proposed a novel attention method for our embedding. Different from SE networks which attention calculated based on channel information and dynamic convolution that attention computed based on average pooling results of input data, we calculate attention directly based on input data.…”

Section: ) Dynamic Saptial Attentionmentioning

confidence: 99%

A Dynamic Spatial-Temporal Attention-Based GRU Model With Healthy Features for State-of-Health Estimation of Lithium-Ion Batteries

Cui

Joe

2021

IEEE Access

View full text Add to dashboard Cite

A proper battery management system (BMS) plays a vital role in ensuring the safety and reliability of electric vehicles (EVs) and other electronic products. Accurate State-of-Health (SOH) estimation of Lithium-ion (Li-ion) batteries is a key factor in a BMS. It is difficult to determine SOH because of the complexity of the electrochemical reactions within the battery. To improve the accuracy of SOH estimation, a dynamic spatial-temporal attention-based gated recurrent unit (DSTA-GRU) model is proposed in this paper. First, we extract six features from the battery's charging and discharging processes that can reflect the aging degree of the battery to some extent. Second, this paper proposes a model to combine spatial attention and temporal attention that can not only consider the effects of states at different time step on the results, but also consider the effects of different features in the space domain. Third, the proposed model is trained and tested on NASA battery datasets and compared with other conventional models. Experiments carried on these data sets demonstrate that our model achieves higher accuracy than other conventional models. INDEX TERMS Lithium-ion battery, state of health, dynamic spatial attention, temporal attention, gated recurrent unit

show abstract

“…Light-weight CNN has performance degradation when suffers computation constrain. Dynamic convolution [17] is presented to overcome this issue, which improves the model performance without increasing the network depth. The general architecture shows in Fig.…”

Section: B Dynamic Convolutionmentioning

confidence: 99%

“…The ResNet and other networks have achieved many successful results in the image field. Nevertheless, to improve the ability to capture the feature and deduce the reduction of the convolution network, our model use Dynamic Convolution [17] to replace the conventional convolution layer. It can dynamically yield the convolution kernel depending on inputs by utilizing the attentions [18].…”

Section: Introductionmentioning

confidence: 99%

Hierarchically Structured Network With Attention Convolution and Embedding Propagation for Imbalanced Few-Shot Learning

et al. 2021

View full text Add to dashboard Cite

Generally, a few-shot distribution shift will lead to a poor generalization. Furthermore, while the number of instances of each class in the real world may significantly different, the existing few-shot classification methods are based on the assumption that the number of samples in each class is equal, which causes the trained classifier invalid. Moreover, through ResNet and WRN (Wide Residual Network) have achieved great success in the image processing field, the depth and width of CNNs constrain the conventional convolution layer performance. Thus, to overcome the above problems, the model of this paper proposes a novel few-shot classification model that uses learning balance variables to decide how much to learn from the imbalance dataset, which dynamically generates the convolution kernel based on each input. In our model, to extend the decision boundaries and enhance the class representations, this paper uses embedding propagation as a regularizer for manifold smoothing. Manifold smoothing can effectively solve the above problems of transductive learning. The interpolations between neural network features based on similarity graphs are used by embedding propagation. Experiments show that embedding propagation can produce a better embedding manifold and our model in standard few-shot datasets, such as miniImagenet, tieredImagenet, CUB has state-of-the-art results. It significantly outperforms the existing few-shot approaches, which consistently improves the accuracy of the models by about 11%.

show abstract

Dynamic Convolution: Attention Over Convolution Kernels

Cited by 615 publications

References 19 publications

HDFNet: Hierarchical Dynamic Fusion Network for Change Detection in Optical Aerial Images

HDFNet: Hierarchical Dynamic Fusion Network for Change Detection in Optical Aerial Images

A Dynamic Spatial-Temporal Attention-Based GRU Model With Healthy Features for State-of-Health Estimation of Lithium-Ion Batteries

Hierarchically Structured Network With Attention Convolution and Embedding Propagation for Imbalanced Few-Shot Learning

Contact Info

Product

Resources

About