• No results found

Deep Learning based Automatic Multi-Class Wild Pest Monitoring Approach using Hybrid Global and Local Activated Features with Stationary Trap Devices

N/A
N/A
Info
Download
Protected

Academic year: 2021

Share "Deep Learning based Automatic Multi-Class Wild Pest Monitoring Approach using Hybrid Global and Local Activated Features with Stationary Trap Devices"

Copied!
11
0
0

Loading.... (view fulltext now)

Download now ( 11 Page )

Full text

Loading

Figure

TABLE 1. Statistics on Two Subsets for our dataset with training subset and validation subset
Fig. 2. Pest monitoring equipment in our work
Fig. 3. Structure of Global activated Feature Pyramid Network (GaFPN)
TABLE 2. Pest Localization Results AP L CNN Backbone Method AP L Inception Faster RCNN 74.99% FPN 76.65% Ours 79.34% ResNet50 Faster RCNN 78.74% FPN 80.29% Ours 82.67%
+2

References

Download now ( PDF - 11 Page - 1.59 MB )

Related documents