Doubling down: sparse grounding with an additional, almost-matching caption for detection-oriented multimodal pretraining.
Giacomo NebbiaAdriana KovashkaPublished in: CVPR Workshops (2022)
Keyphrases
- false alarms
- matching algorithm
- detection algorithm
- automatic detection
- object detection
- detection method
- detection rate
- image matching
- visual features
- sparse representation
- bounding box
- multimedia
- false positives
- anomaly detection
- detection accuracy
- image classification
- high dimensional
- sparse data
- matching process
- graph matching
- template matching
- pattern matching
- multi modal
- object recognition