An Enhanced Encoder-Decoder Network Architecture for Reducing Information Loss in Image Semantic Segmentation.
Zijun GaoQi WangTaiyuan MeiXiaohan ChengYun ZiHaowei YangPublished in: CoRR (2024)
Keyphrases
- semantic segmentation
- network architecture
- information loss
- decoding process
- superpixels
- input image
- image data
- single image
- multiscale
- conditional random fields
- image segmentation
- pixel level
- image content
- neural network
- image features
- image retrieval
- image representation
- scene classification
- weakly supervised
- neural network model
- image set
- object class
- high resolution
- object recognition
- data quality
- data sets
- image classification
- training set
- probabilistic model
- object classes
- object categories
- image matching
- motion estimation
- image regions
- object detection