Hi, Team.
I trained a SSD Caffemodel to detect the targets on the road.
The test results on the PC are basically normal, as shown in Figure 1. However, when the model is converted with TDA2XX and then tested, the results are abnormal, as shown in Figure 2. In Figure 2, the red box is the ground truth, and the green box is the detection result. The result box of the class obstacle at the top right is very wide.
Could you please help me with this problem? Thanks.