Efficient deformable convnets: Rethinking dynamic and sparse operator for vision applications

Y Xiong, Z Li, Y Chen, F Wang, X Zhu… - … Computer Vision …, 2024 - openaccess.thecvf.com
… To overcome these challenges, we propose Deformable … the sparse DCN operator for
practical efficiency. DCNv4 comes with a much faster implementation and an improved operator

[PDF][PDF] Efficient Deformable ConvNets: Rethinking Dynamic and Sparse Operator for Vision Applications Supplementary Material

APL AP APS APM - openaccess.thecvf.com
Environment: We use an A100 80GB SXM GPU to benchmark throughput on all experiments.
The software environment is PyTorch 1.13, CUDA 11.7, cuDNN 8.5. When testing Flash …

Dat++: Spatially dynamic vision transformer with deformable attention

Z Xia, X Pan, S Song, LE Li, G Huang - arXiv preprint arXiv:2309.01430, 2023 - arxiv.org
… visual tokens, prohibiting the applications on high-resolution … to various local modeling
operators, including different types … deformed locations of sparse keys could achieve an efficient

Internimage: Exploring large-scale vision foundation models with deformable convolutions

W Wang, J Dai, Z Chen, Z Huang, Z Li… - … on computer vision …, 2023 - openaccess.thecvf.com
… , so that our model not only has the large effective receptive … the core operator of InternImage
is a dynamic sparse convolution … Revisiting DCNv2. A straightforward way to bridge the gap …

Sdcnet: spatially-adaptive deformable convolution networks for hr nonhomogeneous dehazing

Y Liu, X Wang, Y Zhu, X Fu… - … on Computer Vision and …, 2024 - openaccess.thecvf.com
Efficient deformable convnets: Rethinking dynamic and … Efficient deformable convnets:
Rethinking dynamic and sparse operator for vision applications. arXiv preprint arXiv:2401.06197, …

Vision-rwkv: Efficient and scalable visual perception with rwkv-like architectures

Y Duan, W Wang, Z Chen, X Zhu, L Lu, T Lu… - arXiv preprint arXiv …, 2024 - arxiv.org
… Similar to the Vision Transformer (ViT), our model is designed to efficiently handle … Efficient
deformable convnets: Rethinking dynamic and sparse operator for vision applications. arXiv …

Open panoramic segmentation

J Zheng, R Liu, Y Chen, K Peng, C Wu, K Yang… - … on Computer Vision, 2025 - Springer
… images in computer vision applications are apparent … two critical purposes: (1) efficiently
adapting the frozen CLIP model to … To provide a detailed rethinking process of the deformable

DDEYOLOv9: network for detecting and counting abnormal fish behaviors in complex water environments

Y Li, Z Hu, Y Zhang, J Liu, W Tu, H Yu - Fishes, 2024 - mdpi.com
… By introducing the self-attention mechanism and redefining the … This study improved the
deformable convolution part of the … DCNv4 is an efficient dynamic sparse operator that uses

Computation-efficient deep learning for computer vision: A survey

Y Wang, Y Han, C Wang, S Song… - Cybernetics and …, 2024 - ieeexplore.ieee.org
… recently, some works start to rethink the limitations of the split… computer vision which are
ubiquitous in real-world applications. … for processing voxels efficiently is sparse convolution[201, …

Demystify transformers & convolutions in modern image deep networks

X Hu, M Shi, W Wang, S Wu, L Xing… - … on Pattern Analysis …, 2024 - ieeexplore.ieee.org
… In this comparison, we adopt the deformable convolution v3 (… due to its sparse and binary
nature, which limits information … AI and data-centric AI, particularly exploring their applications