解读flow-guided feature aggregation for video object detection
2017-11-30 20:03
435 查看
文章主要贡献点:
Flow-guided feature aggregation, an end-to-end framework for
video object detection.
Improve the per-frame features by aggregation of nearby features along the motion path, and thus improve the video recognition accuracy.
Or improve the per-frame feature learning by temporal aggregation
数据库:ImageNet VID dataset
3862 video snippet from the traning set
555 snippets from the validation set
Fully annotated
30 object categories (a subset of the categories in the ImageNet DET dataset),
相关工作:
本文工作:
1. the feature extraction network is applied on individual frames to produce the per-frame feature maps
2. To enhance the features at a reference frame, an optical flow network [flownet] estimates themotions between the nearby frames an the reference frame.
3. The feature maps from nearby frames are warped to the reference maps, as well as its own feature maps on the reference frame, areaggregated according to an adaptive weighting network.
4. The resulting aggregated feature maps are then fed to the detection network to produce the detection result on the reference frame.
System: Feature extraction + flow estimation + feature aggregation + detection
Flow-guided feature aggregation, an end-to-end framework for
video object detection.
Improve the per-frame features by aggregation of nearby features along the motion path, and thus improve the video recognition accuracy.
Or improve the per-frame feature learning by temporal aggregation
数据库:ImageNet VID dataset
3862 video snippet from the traning set
555 snippets from the validation set
Fully annotated
30 object categories (a subset of the categories in the ImageNet DET dataset),
相关工作:
本文工作:
1. the feature extraction network is applied on individual frames to produce the per-frame feature maps
2. To enhance the features at a reference frame, an optical flow network [flownet] estimates themotions between the nearby frames an the reference frame.
3. The feature maps from nearby frames are warped to the reference maps, as well as its own feature maps on the reference frame, areaggregated according to an adaptive weighting network.
4. The resulting aggregated feature maps are then fed to the detection network to produce the detection result on the reference frame.
System: Feature extraction + flow estimation + feature aggregation + detection
相关文章推荐
- 【论文笔记】视频物体检测(VID)系列 FGFA:Flow-Guided Feature Aggregation for Video Object Detection
- 视频目标检测--Flow-Guided Feature Aggregation for Video Object Detection
- Object Detection -- 论文FPN(Feature Pyramid Networks for Object Detection)解读
- 论文解读之Feature Pyramid Networks for Object Detection
- 【论文笔记】Rich feature hierarchies for accurate object detection and semantic segmentation
- 行人检测论文笔记:Fast Feature Pyramids for Object Detection?
- 论文笔记 《Rich feature hierarchies for accurate object detection and semantic segmentation》
- [深度学习论文笔记][Object Detection] Rich feature hierarchies for accurate object detection and semantic seg
- [论文解读] MSCNN: A Unified Multi-scale Deep Convolutional Neural Network for Fast Object Detection
- 论文笔记——Rich feature hierarchies for accurate object detection and semantic segmentation
- FPN Feature Pyramid Network for Object Detection
- Rich feature hierarchies for accurate object detection and semantic segmentation
- Rich feature hierarchies for accurate object detection and semantic segmentation(泛读)
- 论文提要“Fast Feature Pyramids for Object Detection”
- 经典文章系列:Feature Pyramid Networks for Object Detection(FPN)论文阅读
- 论文提要“Fast Feature Pyramids for Object Detection”
- 多尺度R-CNN论文笔记(6): Feature Pyramid Networks for Object Detection
- R-CNN:Rich feature hierarchies for accurate object detection and semantic segmentation(阅读)
- R-CNN论文笔记《Rich feature hierarchical for accurate object detection and semantic segmentation》
- Deep Feature Flow for Video Recognition读书笔记