Expandable-RCNN: toward high-efficiency incremental few-shot object detection.
Yiting LiSichao TianHaiyue ZhuYeying JinKeqing WangJun MaCheng XiangPrahlad VadakkepatPublished in: Frontiers Artif. Intell. (2024)
Keyphrases
- high efficiency
- object detection
- high accuracy
- incremental learning
- real and synthetic datasets
- face detection
- object classification
- video sequences
- software architecture
- detection algorithm
- memory space
- computer vision
- arbitrary shape
- multi class
- scene understanding
- object segmentation
- visual features
- background subtraction
- pedestrian detection
- object recognition
- human detection
- video shots
- incremental version
- object categories
- key frames
- video data
- video content
- result quality
- change detection
- video streams
- space time
- scene recognition
- video indexing
- super resolution reconstruction
- high resolution
- shot change detection