Image to Sphere: Learning Equivariant Features for Efficient Pose Prediction.
David KleeOndrej BizaRobert PlattRobin WaltersPublished in: ICLR (2023)
Keyphrases
- image features
- low level
- image data
- image classification
- learning algorithm
- linear predictors
- extracted features
- visual object recognition
- image regions
- keypoints
- pose estimation
- prior knowledge
- input image
- original images
- invariant features
- extracting features
- feature vectors
- image segmentation
- single image
- global features
- multiscale
- feature extraction
- spatial information
- position and orientation
- feature values
- pose space
- prediction accuracy
- feature points
- edge detection
- fingerprint images
- object detectors
- test images
- vector valued
- image retrieval
- feature extractor
- face recognition
- computer vision