.png){kind=icon}
Return
Learning Multi-Scale Features Using Dilated Convolution for Contour Detection
.png){kind=icon}
OA
AI
DOI：10.1109/ACCESS.2023.3289203.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
For the contour detection task, we use the EfficientNet model as the backbone network and propose a network model that uses dilated convolution for multi-scale optimization. The network is accumulated top-down layer by layer, combining multiple optimization modules concat together to achieve a richer feature representation. To fuse feature information at different scales, we introduce a new Multi-scale optimization module to replace the use of deeper network structures or more complex decoding methods, which uses channel attention module to learn the correlation between channels and then uses dilated convolution of different scales to enhance contextual information. High generalization performance and accuracy are obtained in comparison with recent deep learning-based contour detection models. We evaluate our approach on two datasets, i.e., BSDS500 and NYUD-v2, achieving an ODS F-measure value of 0.828 on BSDS500. In particular, the results of BSDS500 exceed the human-level performance under more stringent criteria.
Keywords:
Contour detection
decode network
deep refinement network
multi-scale integration
Journal
IF:
3.6
Papers:
9.8W
Citations:
29.4W
Organization
No organization information available
Cited Papers
Contour detection model with multi-scale integration based on non-classical receptive field.png)
NEUROCOMPUTING
NEUROCOMPUTINGIF6.5
PREAI
ShareSaveShareSaveThermal comfort, perceived air quality, and cognitive performance when personally controlled air movement is used by tropically acclimatized personsIndoor Air
OAAI
Indoor AirIF0
OAAIShareSaveShareSaveShareSaveShareSave
View more