在特定设备上运行实现tensorflow-lite的android应用程序时出现错误

如何解决在特定设备上运行实现tensorflow-lite的android应用程序时出现错误

在特定设备上运行实现 tensorflow-lite 的 android 应用程序时发生错误。

例如)LG X4、LG Q52

调用函数时出错 -> tfLite.runForMultipleInputsOutputs(inputArray,outputMap);

这是我的错误日志。

E/AndroidRuntime: FATAL EXCEPTION: inference
Process: org.tensorflow.lite.examples.detection,PID: 13469
java.lang.IllegalArgumentException: Internal error: Failed to run on the given Interpreter: Next operations are not supported by GPU delegate:
ADD: 
CONCATENATION: 
CONV_2D: 
EXP: Operation is not supported.
LEAKY_RELU: 
LOGISTIC: 
MAX_POOL_2D: 
MUL: 
RESHAPE: 
RESIZE_BILINEAR: 
SPLIT: Operation is not supported.
SPLIT_V: Operation is not supported.
STRIDED_SLICE: 
SUB: Expected 2 input tensor(s),but node has 1 runtime input(s).
First 61 operations will run on the GPU,and the remaining 144 on the CPU.
OpenCL library not loaded - dlopen fai
    at org.tensorflow.lite.NativeInterpreterWrapper.run(Native Method)
    at org.tensorflow.lite.NativeInterpreterWrapper.run(NativeInterpreterWrapper.java:154)
    at org.tensorflow.lite.Interpreter.runForMultipleInputsOutputs(Interpreter.java:314)
    at org.tensorflow.lite.examples.detection.tflite.YoloV4Classifier.getDetectionsForTiny(YoloV4Classifier.java:421)
    at org.tensorflow.lite.examples.detection.tflite.YoloV4Classifier.recognizeImage(YoloV4Classifier.java:461)
    at org.tensorflow.lite.examples.detection.tflite.YoloV4Classifier.recognizeImage(YoloV4Classifier.java:50)
    at org.tensorflow.lite.examples.detection.DetectorActivity$3.run(DetectorActivity.java:304)
    at android.os.Handler.handleCallback(Handler.java:883)
    at android.os.Handler.dispatchMessage(Handler.java:100)
    at android.os.Looper.loop(Looper.java:214)
    at android.os.HandlerThread.run(HandlerThread.java:67)
I/Process: Sending signal. PID: 13469 SIG: 9
Connected to process 13858 on device 'lge-lm_q520n-LMQ520NOFAIMNHYEMS'.

请帮帮我

解决方法

您没有提供其他详细信息，如果它适用于其他设备或模型是如何生成的。我以前遇到过这个错误。它可能与设备无关。根据错误日志，您似乎是 using this repository 使用 yolov4 tflite 模型进行对象检测任务。

java.lang.IllegalArgumentException: Internal error: Failed to run on the given Interpreter: Next operations are not supported by GPU delegate:
EXP: Operation is not supported.
SPLIT: Operation is not supported.
SPLIT_V: Operation is not supported.

在将保存的模型转换为 tflite 时，在 tf.lite.OpsSet.SELECT_TF_OPS 中使用 converter.target_spec.supported_ops。同样在 build.gradle 中使用以下任一项，

implementation 'org.tensorflow:tensorflow-lite-select-tf-ops:2.3.0'

implementation 'org.tensorflow:tensorflow-lite-select-tf-ops:0.0.0-nightly'。

这会增加 apk 的大小，但这可以通过 tensorflow-lite-select-tf-ops.aar 的 selective build 解决。