Anchor free object detection with mask attention
The anchor-free method based on key point detection has made great progress. However, the anchor-free method is too dependent on using a convolutional network to generate a rough heat map. This is difficult to detect for objects with a large size variation and dense and overlapping objects. To solve this problem, first, we propose a mask attention mechanism for object detection methods. And make full use of the advantages of the attention mechanism to improve the accuracy of network detection heat map generation. Then, we designed an optimized fire model to reduce the size of the model. The fire model is an extension of grouped convolution. The fire model allows each group of convolutional network features to learn the same feature through purposeful grouping. In this paper, the mask attention mechanism uses object segmentation images to guide the generation of corner heat maps. Our approach achieved an accuracy of 91.84% and a recall 89.83% in the Tencent-100K dataset. Compared with the popular object detection methods, the proposed method has advantages in model size and accuracy.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Due to technical limitations, full-text HTML conversion of this manuscript could not be completed. However, the manuscript can be downloaded and accessed as a PDF.
Posted 03 Jun, 2020
On 14 Jul, 2020
On 03 Jul, 2020
Received 25 Jun, 2020
Received 25 Jun, 2020
Received 18 Jun, 2020
Received 18 Jun, 2020
On 04 Jun, 2020
Invitations sent on 31 May, 2020
On 31 May, 2020
On 31 May, 2020
On 31 May, 2020
On 28 May, 2020
On 27 May, 2020
On 27 May, 2020
Received 26 Apr, 2020
On 26 Apr, 2020
Received 25 Apr, 2020
On 04 Apr, 2020
On 01 Apr, 2020
On 31 Mar, 2020
On 31 Mar, 2020
On 31 Mar, 2020
Received 30 Mar, 2020
On 28 Mar, 2020
On 28 Mar, 2020
On 27 Mar, 2020
On 27 Mar, 2020
Received 27 Mar, 2020
Invitations sent on 24 Mar, 2020
On 24 Mar, 2020
On 27 Feb, 2020
On 26 Feb, 2020
On 26 Feb, 2020
On 24 Feb, 2020
Anchor free object detection with mask attention
Posted 03 Jun, 2020
On 14 Jul, 2020
On 03 Jul, 2020
Received 25 Jun, 2020
Received 25 Jun, 2020
Received 18 Jun, 2020
Received 18 Jun, 2020
On 04 Jun, 2020
Invitations sent on 31 May, 2020
On 31 May, 2020
On 31 May, 2020
On 31 May, 2020
On 28 May, 2020
On 27 May, 2020
On 27 May, 2020
Received 26 Apr, 2020
On 26 Apr, 2020
Received 25 Apr, 2020
On 04 Apr, 2020
On 01 Apr, 2020
On 31 Mar, 2020
On 31 Mar, 2020
On 31 Mar, 2020
Received 30 Mar, 2020
On 28 Mar, 2020
On 28 Mar, 2020
On 27 Mar, 2020
On 27 Mar, 2020
Received 27 Mar, 2020
Invitations sent on 24 Mar, 2020
On 24 Mar, 2020
On 27 Feb, 2020
On 26 Feb, 2020
On 26 Feb, 2020
On 24 Feb, 2020
The anchor-free method based on key point detection has made great progress. However, the anchor-free method is too dependent on using a convolutional network to generate a rough heat map. This is difficult to detect for objects with a large size variation and dense and overlapping objects. To solve this problem, first, we propose a mask attention mechanism for object detection methods. And make full use of the advantages of the attention mechanism to improve the accuracy of network detection heat map generation. Then, we designed an optimized fire model to reduce the size of the model. The fire model is an extension of grouped convolution. The fire model allows each group of convolutional network features to learn the same feature through purposeful grouping. In this paper, the mask attention mechanism uses object segmentation images to guide the generation of corner heat maps. Our approach achieved an accuracy of 91.84% and a recall 89.83% in the Tencent-100K dataset. Compared with the popular object detection methods, the proposed method has advantages in model size and accuracy.
Figure 1
Figure 2
Figure 3
Figure 4
Figure 5
Figure 6
Figure 7
Figure 8
Figure 9
Due to technical limitations, full-text HTML conversion of this manuscript could not be completed. However, the manuscript can be downloaded and accessed as a PDF.