You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I made a Pull request to compile minimal example in Dev Board #97 (comment)
But I can't execute it successfully with edgetpu models.
mendel@undefined-dog:~/coral/edgetpu/src/cpp/examples$ LD_LIBRARY_PATH=/home/mendel/coral/edgetpu/libedgetpu/direct/aarch64 ./minimal ~/coral/edgetpu/test_data/mobilenet_v1_1.0_224_quant_edgetpu.tflite ~/coral/edgetpu/test_data/resized_cat.bmp
ERROR: Internal: Unsupported data type in custom op handler: -610837120
ERROR: Node number 0 (edgetpu-custom-op) failed to prepare.
Failed to allocate tensors.
Segmentation fault
But works with tflite models
LD_LIBRARY_PATH=/home/mendel/coral/edgetpu/libedgetpu/direct/aarch64 ./minimal
~/coral/edgetpu/test_data/mobilenet_v1_1.0_224_quant.tflite ~/coral/edgetpu/test_data/resized_cat.bmp
[Image analysis] max value index: 286 value: 0.804688
The models works with python demos
mendel@undefined-dog:~/coral/tflite/python/examples/classification$ LD_LIBRARY_PATH=/home/mendel/coral/edgetpu/libedgetpu/direct/aarch64 python3 classify_image.py --model ~/coral/edgetpu/test_data/mobilenet_v1_1.0_224_quant_edgetpu.tflite --input ~/coral/edgetpu/test_data/resized_cat.bmp
----INFERENCE TIME----
Note: The first inference on Edge TPU is slow because it includes loading the model into Edge TPU memory.
14.6ms
2.3ms
2.3ms
2.3ms
2.3ms
-------RESULTS--------
286: 0.79297
The text was updated successfully, but these errors were encountered:
@jlz3008 Really appreciate the PR and the attention for this repo!!
Unfortunately, our repo right now has some blocking issues to take in PRs due to how it's structured internally. We are working diligently to come up with a solution for PRs which I expected to be working in the future.
On another note, your build seems to be segfault-ing on the dev board, which isn't something we'd need to add. I recently worked on a native build solution to allow this example to be build natively, tested on my laptop, dev board, and my pi4. Please checkout this minimal repo: https://github.com/Namburger/edgetpu-minimal-example
Let me know if you have issues trying it, here are instructions to build & run on the dev board:
Thanks for your code @Namburger . It works.
I can see that the minimal.cc and model_utils.cc code is almost identical to google-coral/edgetpu repo so I guess the difference is in the tflite and other dependences repos.
I need some time to understand how to the build works in your repo code.
I made a Pull request to compile minimal example in Dev Board #97 (comment)
But I can't execute it successfully with edgetpu models.
But works with tflite models
The models works with python demos
The text was updated successfully, but these errors were encountered: