A Hybrid RNN-HMM Approach for Weakly Supervised Temporal Action Segmentation.
Hilde KuehneAlexander RichardJuergen GallPublished in: CoRR (2019)
Keyphrases
- weakly supervised
- semantic segmentation
- superpixels
- fully supervised
- segmentation algorithm
- hidden markov models
- object class
- segmentation method
- topic models
- object segmentation
- relation extraction
- semi supervised
- nearest neighbor
- image segmentation
- conditional random fields
- level set
- medical images
- multiscale
- named entities
- computer vision
- bounding box
- image regions
- multiple objects
- generative model
- object detectors
- natural images
- pairwise