Hello, I want to train a chip defect detection model using tensorflow objection detection module and run it on opencv of c++ version by generating .pb files with python and call it with opencv of cpp version. But the problem is that the detection output of opencv dnn is different from the detection output of tennsorflow object detection module.
Environment: 1. tensorflow object detection module tf 1.13.0 2. opencv dnn 4.2.0 opencl (temporarily using cpu, not cuda) 3. ssd_mobilenet_v1_coco_2017_11_17.pbtxt 4. ssd_mobilenet_v2_coco.config 5. having gotten rid of the input difference