TensorFlow Lite 应用程序

Question

如何将自定义模型应用于 tensorflowlite？ 在指令中，我想将模型粘贴到资产中并删除下载模型 gradle，但是当它被应用时，应用程序简单粉碎，有什么问题？

已编辑：

我在 tensorflow lite 中使用自定义训练时遵循说明，但在将 api 模型文件更改为我的模型并评论下载模型后，应用程序简单崩溃。我尝试使用 microsoft general compact 和 compact[s1] 的自定义 ai 并按照指令https://medium.com/datadriveninvestor/how-to-train-your-own-custom-model-with-tensorflow-object-detection-api-and-deploy-it-into-android-aeacab7fa76f 两者都不起作用，compact[s1] 使应用程序混乱和崩溃，而 compact 完全崩溃所以，有什么问题，自定义 AI 简单不能在 tensorflowlite 中使用吗？

这个东西以调试模式显示：

    java.lang.IllegalArgumentException: Internal error: Failed to apply delegate: Regular TensorFlow ops are not supported by this interpreter. Make sure you apply/link the Flex delegate before inference.
    Node number 328 (FlexNonMaxSuppressionV4) failed to prepare.

        at org.tensorflow.lite.NativeInterpreterWrapper.applyDelegate(Native Method)
        at org.tensorflow.lite.NativeInterpreterWrapper.applyDelegates(NativeInterpreterWrapper.java:324)
        at org.tensorflow.lite.NativeInterpreterWrapper.init(NativeInterpreterWrapper.java:82)
        at org.tensorflow.lite.NativeInterpreterWrapper.<init>(NativeInterpreterWrapper.java:63)
        at org.tensorflow.lite.Interpreter.<init>(Interpreter.java:237)
        at org.tensorflow.lite.examples.detection.tflite.TFLiteObjectDetectionAPIModel.create(TFLiteObjectDetectionAPIModel.java:140)
        at org.tensorflow.lite.examples.detection.DetectorActivity.onPreviewSizeChosen(DetectorActivity.java:112)
        at org.tensorflow.lite.examples.detection.CameraActivity.onPreviewFrame(CameraActivity.java:200)
        at android.hardware.Camera$EventHandler.handleMessage(Camera.java:1222)
        at android.os.Handler.dispatchMessage(Handler.java:102)
        at android.os.Looper.loop(Looper.java:154)
        at android.app.ActivityThread.main(ActivityThread.java:6316)
        at java.lang.reflect.Method.invoke(Native Method)
        at com.android.internal.os.ZygoteInit$MethodAndArgsCaller.run(ZygoteInit.java:872)
        at com.android.internal.os.ZygoteInit.main(ZygoteInit.java:762)```

Edited v2:
Yes, it show up in netron.
[![this is the image of the other persons model][1]][1]
[![this one is my model in microsoft compact[s1][2]][2]
the other model of microsoft compact is to big for png so, it was saved as svg don't know how to show it here.

[![here is the properties of compact[s1][3]][3]
[![here is the properties of compact][4]][4]
[![here is the properties of other person][5]][5]


  [1]: https://i.stack.imgur.com/BtkZ1.jpg
  [2]: https://i.stack.imgur.com/ffnOU.png
  [3]: https://i.stack.imgur.com/mx7dx.png
  [4]: https://i.stack.imgur.com/AyXrt.png
  [5]: https://i.stack.imgur.com/K1odl.png

Answer 1

TensorFlow Lite 不支持您的模型使用的运算符。这是supported ops的列表。

您可以查看如何创建model easily convertible to TensorFlow Lite。