Question about Fpn-SSD-MobileNet-v2 Postprocess

[June1124]

I trained the object detection model using fpn-ssd-mobilenetv2 inside eIQ and successfully exported it in tflite format, but its output is float32[1,19125,7], I expect the result to be TFLITE_DETECTION_POSTPROCESS, which has the dimensions [1,1,4]. . Is there a solution for this, please? Or how to decode [1,19125,7]?

My eIQ Portal version is 2.5.12.

[david_piskula]

Hello June, why do you expect the result to be TFLITE_DETECTION_POSTPROCESS and where does this variable/macro come from? Can you explain your issue in more detail, please?