When i trained the ssd_mobilenetv2_fpnlite_320x320, and save it as saved_model.pb, it works fine, the bounding box appear with the label of the a bill, but when i try to deploy it in the latest version of Tensorflow object detection sample app, there is no label, only the bounding box appear when i place a bill in front of the camera, but some bills are recognize. I converted the .pb model into tflite with metadata. I think that there could be a problem in converting the model into tflite format.
this is how i did the conversion
