효율적인 객체 검출을 위해 Attention Process를 적용한 경량화 모델에 대한 연구

목차 (Table of Contents)

• 요약
• Abstract
• 1. 서론
• 2. 관련연구
• 2.1 Feature pyramid
• 2.2 SSD
• 3. 제안하는 방법
• 3.1 Convolution-Attention block
• 3.2 Lightweight feature attention process
• 3.3 Loss function
• 4. 실험 및 결과
• 4.1 PASCAL VOC 2007
• 4.2 Network Inference Time
• 5. 결론
• REFERENCES

참고문헌 (Reference)

1 Redmon, J., "You only look once: Unified, real-time object detection" 779-788, 2016

2 Redmon, J., "Yolov3: An incremental improvement"

3 Redmon, J., "YOLO9000: better, faster, stronger" 7263-7271, 2017

4 Chollet, F., "Xception: Deep learning with depthwise separable convolutions" 1251-1258, 2017

5 Liu, W, "Ssd: Single shot multibox detector" Springer 21-37, 2016

6 He, K, "Spatial pyramid pooling in deep convolutional networks for visual recognition" 37 (37): 1904-1916, 2015

7 Zhang, S., "Single-shot refinement neural network for object detection" 4203-4212, 2018

8 Girshick, R, "Rich feature hierarchies for accurate object detection and semantic segmentation" 580-587, 2014

9 Lin, M., "Network in network"

10 Howard, A. G, "Mobilenets: Efficient convolutional neural networks for mobile vision applications"

11 Cao, J., "Hierarchical shot detector" 9705-9714, 2019

12 Lin, T. Y., "Feature pyramid networks for object detection" 2117-2125, 2017

13 Ren, S, "Faster r-cnn: Towards real-time object detection with region proposal networks"

14 Tan, M., "Efficientnet: Rethinking model scaling for convolutional neural networks" PMLR 6105-6114, 2019

15 Neubeck, A., "Efficient non-maximum suppression" IEEE 3 : 850-855, 2006

16 Fu, C. Y, "Dssd: Deconvolutional single shot detector"