Enhancement of SSD by concatenating feature maps for object detection

By Jisoo Jeong, Hyojin Park, Nojun Kwak. 2017

336 citations recorded by the source corpus.

We propose an object detection method that improves the accuracy of the conventional SSD (Single Shot Multibox Detector), which is one of the top object detection algorithms in both aspects of accuracy and speed. The performance of a deep network is known to be improved as the number of feature maps increases. However, it is difficult to improve the performance by simply raising the number of feature maps. In this paper, we propose and analyze how to use feature maps effectively to improve the performance of the conventional SSD. The enhanced performance was obtained by changing the structure close to the classifier network, rather than growing layers close to the input data, e.g., by replacing VGGNet with ResNet. The proposed network is suitable for sharing the weights in the classifier networks, by which property, the training can be faster with better generalization power. For the Pascal VOC 2007 test set trained with VOC 2007 and VOC 2012 training sets, the proposed network with the input size of 300 x 300 achieved 78.5% mAP (mean average precision) at the speed of 35.0 FPS (frame per second), while the network with a 512 x 512 sized input achieved 80.8% mAP at 16.6 FPS using Nvidia Titan X GPU. The proposed network shows state-of-the-art mAP, which is better than those of the conventional SSD, YOLO, Faster-RCNN and RFCN. Also, it is faster than Faster-RCNN and RFCN.

Subfields: Computer Networks and Communications, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

1. Fuchun Sun
2. Jieping Ye
3. L. Jiao
4. Rong Qu
5. Zuoxin Li
6. Fuqiang Zhou
7. Jianlin Wang
8. Bin Liu

In Computer Vision and Pattern Recognition

1. In-So Kweon
2. Zhenwei Shi
3. Sanghyun Woo
4. Huaping Liu
5. Keyan Chen
6. Tao Kong
7. Zhengxia Zou
8. Fuchun Sun

Citations in the corpus

• StairNet: Top-Down Semantic Aggregation for Accurate One Shot Detection (2017, background): “7 shows the results that even in the high-recall range our model achieves high-precision and outperforms SSD and R-SSD [15].”
• Deep Feature Pyramid Reconfiguration for Object Detection (2018, background): “Several works are trying to develop more suitable connection manners [25,46,48], or to add more operations before combination [28].”
• Surface Defect Detection Methods for Industrial Products: A Review (2021, background): “Literature [95] integrated the features of different layers through rainbow concatenation (pooling and deconvolution), increasing the number of feature maps in different layers while increasing the feature map relationship between different layers, which solves the problem of small goals to a certai”
• Real-Time Detection of Apple Leaf Diseases Using Deep Learning Approach Based on Improved Convolutional Neural Networks (2019, methodology): “The methodology of Rainbow concatenation, which is implemented in R-SSD [26], is applied to further improve the detection accuracy for small objects that correspond to apple”
• A Survey of Deep Learning-Based Object Detection (2019, background): “Some works [93][94][95] improve detection accuracy on the basis of [8].”
• FSSD: Feature Fusion Single Shot Multibox Detector (2017, methodology): “RSSD300 [16] 07+12 ! VGGNet 35 Titan X 8732 300× 300 78.”
• Object Detection in 20 Years: A Survey (2019, background): “Bottom-up fusion feeds forward shallow features to deeper layers via skip connections [237, 240–242].”
• DC-SPP-YOLO: Dense Connection and Spatial Pyramid Pooling Based YOLO for Object Detection (2019, methodology): “[23] proposed different methods of fusing the multi-scale feature maps to improve SSD and got better performances than SSD, for which the finer layers of the networks could utilize the contextual information learned from the coarser layers of the networks.”
• Parallel Feature Pyramid Network for Object Detection (2018, methodology): “Incorporating context information at different scales can facilitate several visual classification tasks [1,13,18,28].”
• Efficient Featurized Image Pyramid Network for Single Shot Detector (2019, neutral): “…optimal trade-off between detection accuracy and speed by providing a Methods Backbone input size mAP FPS SSD [28] VGG-16 300 × 300 77.2 120 R-SSD [20] VGG-16 300 × 300 78.5 35 RUN [23] VGG-16 300 × 300 79.2 40 ESSD [35] VGG-16 300 × 300 79.4 25 DSSD [12] ResNet-101 321 × 321 78.6 9.5 DES [34]…”