Hi. I am the student who asked the same question for the last time.
Two days ago, I faced the problem of reasoning of two similar models, and I tried to solve it.
However, the problem has not been solved even though we have fixed many suspicious parts including extended USB cable.
I tried to buy one more Raspberry PI and run the model but it did not work out.
If possible, I would appreciate it if you could look at the structure and conversion process of my two models and what parts are wrong.
The first is the structure of a well-driven model and the IR conversion process using mo_tf.py.
Successful Case Markdown Link
Failure Case Markdown Link
No matter how I find it, I can not see the difference between the two cases.
If possible, I would like to hear from Intel.
I will attach IR files and test images for both models.
I would like you to test the two models.
For more complete information about compiler optimizations, see our Optimization Notice.