Closed talcs closed 2 years ago
Hi, in this issue you are trying with cpu tflite model. Can you please try with mobilenet_v2_1.0_224_inat_bird_quant_edgetpu.tflite
.
Wow, thanks! I'm pretty shocked... wasn't aware of the fact that there were two files with similar names.
The updated output:
C:\programs\coral\pycoral>python .\examples\classify_image.py --model .\test_data\mobilenet_v2_1.0_224_inat_bird_quant_edgetpu.tflite --labels .\test_data\inat_bird_labels.txt --input .\test_data\parrot.jpg
----INFERENCE TIME----
Note: The first inference on Edge TPU is slow because it includes loading the model into Edge TPU memory.
14.4ms
4.8ms
4.9ms
4.8ms
4.7ms
-------RESULTS--------
Ara macao (Scarlet Macaw): 0.76953
Problem's solved 😃
Description
Hello,
I have a Windows 10 machine and I could only execute the Coral example using the old drivers (edgetpu_runtime_20210119), which caused abnormal runtimes of ~1800ms per image.
Later, I fixed the
install.bat
script in the latest version (edgetpu_runtime_20210726):Before change
After change
Now the
failed to load delegate from edgetpu.dll
error message has disappeared, yet a new error started appearing, where the script hangs up with no error message when it callsinterpreter = make_interpreter(*args.model.split('@'))
.Insights
return tflite.load_delegate(_EDGETPU_SHARED_LIB, options or {})
.libusb-1.0.dll
file of the latest edgetpu_runtime release makes the problems. If I copy the file from the 20210119 version toC:\Windows\System32
, then I fallback to the previous problem of very slow model execution.Any advice?
Click to expand!
### Issue Type Build/Install, Support ### Operating System Windows 10 ### Coral Device USB Accelerator ### Other Devices _No response_ ### Programming Language Python 3.7 ### Relevant Log Output _No response_