Rawgment: Noise-Accounted RAW Augmentation Enables Recognition in a Wide Variety of Environments.
Masakazu YoshimuraJunji OtsukaAtsushi IrieTakeshi OhashiPublished in: CVPR (2023)
Keyphrases
- wide variety
- recognition rate
- recognition accuracy
- object recognition
- real world
- anti noise
- noise level
- feature extraction
- image noise
- signal to noise ratio
- noisy environments
- noise reduction
- additive noise
- noise model
- automatic recognition
- image recognition
- pattern recognition
- recognition process
- visual recognition
- gesture recognition
- noise free
- human activities
- activity recognition
- random noise
- dynamic environments
- unsupervised learning
- input data
- shape recognition
- data sets
- noise removal
- median filter
- high level
- noisy images
- noisy data
- action recognition
- multiscale