'Is there any method to decrease tensorflow lite invoke time in c++?
System information
- windows 10, x64
- TensorFlow version: 2.7.0
- Python version: 3.9.5
- Bazel version : 3.7.2 5.GCC/Compiler version : 11.2.0
I run a custom CNN tflite model on tensorflow lite using c++ where i got invoke time 2.5 second and same tensorflow model version i run on python where i got 1.19 second . As i know c++ is more faster than python then what is the issue ?
Note : because of commercial reason i can't give the code snapshot.
Solution 1:[1]
TFLite focuses more on on-device performance. So it is not as optimized when you run it on a x86 machines compared to arm devices for example.
Sources
This article follows the attribution requirements of Stack Overflow and is licensed under CC BY-SA 3.0.
Source: Stack Overflow
| Solution | Source |
|---|---|
| Solution 1 | Karim Nosseir |
