Fig. 5
From: Adaptive cascade single-shot detector on wireless sensor networks
The results of the bounding box detector demonstrate an imbalance between the classes. Each column represents a comparison of the number of borders for each class using different intersections within a 10% to 90% joint threshold range. SENET and SSD models are used to compare the accuracy of identifying objects of different sizes. The overall performance of SENET model is better than that of contrast model, and the recognition accuracy of chair with sparse structure is much higher than that of SSD