Recently I used image classification in mobile and it works good.Will there be documents to deploy object detection,pose estimation, segmentation in mobile as in tensorflow in future?
I was able to get a segmentation model work by using the classification tutorial from the pytorch homepage. I used <tensor>.getDataAsFloatArray()
to get the output values.
In any case, currently the amount of documentation content for pytorch-mobile is ridiculously low. Like zero tutorials…
Yeah. But I saw in discussion saying that they are looking into it. Maybe in near future we can except those.
did you get it out as a single dimensional array and have to put it back into it proper structure?
I only tried image classification. But wanted to do for other tasks also. If torch has these tutorials then I can fully move into torch
I have been using a mobilenet-ssd successfully.
I use the below function to restructure the array.
public float[][] floatArray1dto2d(final float[] array, final int rows, final int cols) {
if (array.length != (rows * cols))
throw new IllegalArgumentException("Invalid array length");
float[][] outArray = new float[rows][cols];
for (int i = 0; i < rows; i++)
System.arraycopy(array, (i * cols), outArray[i], 0, cols);
return outArray;
}
@Redney_Quartz what’s the size of your model? and on which platform you was able to use the model (ios/android) ?