Mastering Computer Vision with TensorFlow 2.x

TensorFlow Object Detection API – tflite_convert

In the following code, we are defining the location of the frozen model .pb file and the corresponding tflite file. Then, we're resizing the input color image to (300,300) in shape for each of the three RGB axes and converting the image into a normalized tensor, which then becomes the input array for the conversion. There are four output arrays, all of which are defined as follows:

• TFLite_Detection_PostProcess—detection boxes
• TFLite_Detection_PostProcess:1—detection classes
• TFLite_Detection_PostProcess:2—detection scores
• TFLite_Detection_PostProcess:3—number of detections

!tflite_convert \--graph_def_file=/content/models/research/fine_tuned_model/tflite_graph.pb \--output_file=/content/models/research/fine_tuned_model/burgerfries.tflite ...