How to use EfficientNet on ROCK Pi N10

DavideMiceli · December 30, 2021, 5:33pm

Hi,
i’ve tried to use EfficientNetB0 for image classification on Rock Pi N10 but the network doesn’t work properly.
I downloaded efficientnet_lite0_int8_2.tflite from TensorFlow Hub:
https://tfhub.dev/tensorflow/lite-model/efficientnet/lite0/int8/2
Then i ran this test i found on rknn-toolkit repository:

github.com

rockchip-linux/rknn-toolkit/blob/master/examples/tflite/mobilenet_v1/test.py

import numpy as np
import cv2
from rknn.api import RKNN


def show_outputs(outputs):
    output = outputs[0][0]
    output_sorted = sorted(output, reverse=True)
    top5_str = 'mobilenet_v1\n-----TOP 5-----\n'
    for i in range(5):
        value = output_sorted[i]
        index = np.where(output == value)
        for j in range(len(index)):
            if (i + j) >= 5:
                break
            if value > 0:
                topi = '{}: {}\n'.format(index[j], value)
            else:
                topi = '-1: 0.0\n'
            top5_str += topi

This file has been truncated. show original

using the efficientnet instead of mobilenet.
I also tried to do some changes on the config part of the script and in the preprocessing of the image but none of them worked.
I have:

Python 3.7.3
rknn-toolkit 1.4.0
Does anyone know what i did wrong or where i can find an example test for EfficientNet?
Thank you a lot for the help.