Efficient Training of 3D Seismic Image Fault Segmentation Network under Sparse Labels by Weakening Anomaly Annotation.
Yimin DouKewen LiJianbing ZhuTiming LiShaoquan TanZongchao HuangPublished in: CoRR (2021)
Keyphrases
- segmentation method
- segmented images
- segmentation algorithm
- image segmentation
- multiscale
- grey level
- image analysis
- image data
- image regions
- semantic segmentation
- weakly labeled
- labeling process
- auto annotation
- input image
- single image
- test images
- edge detection
- segmentation accuracy
- image pixels
- energy function
- image features
- region segmentation
- image retrieval
- automatic image annotation
- image collections
- network traffic
- image content
- active learning
- training set
- medical images
- homogeneous regions
- annotation effort
- adaptive thresholding
- region growing
- image segments
- image classification
- pixel level
- energy functional
- training examples
- level set
- labeled images
- image representation
- intrusion detection
- segmentation errors
- signal processing
- high resolution
- image processing
- gray value
- bounding box
- visual concepts
- multiple objects
- classifier training
- anomaly detection
- image set
- training data
- active contours
- neural network