How to convert your own custommodel to tensorflow lite

I’m trying to convert my custommodel to tensorflow lite for image recognition, but I’m having trouble recognizing it after the conversion.
The image is recognized when I check it on tensorboard after training it with Colab, so it seems to break down in the process of conversion.
Do you know of any way to convert the data correctly?

I am using tflite_convert as the current conversion method.

The author is Japanese, so I am using a translation tool to write this text. If you have any questions, please feel free to contact me. I apologize for this.

Have you tried to validate it side by side with something like this?