-
This is my project of tensorflow model for microcontroller now it is only sute for tflite microcontroller
-
There are only mobilenetv1 and mobilenetv3 implemented,the converting between mobilenetv3 and mobilennetv3.tflite has a degradation problem you may check this issue to catch problem conversion issue
-
onnx2tflite tool is implicated just find it in forx_onnx/onnx2tflite.py
- Chinese readme document
- merge data gengerate code and train code
- The following test is only test on 2024_05_06_val data_set with qvga image format and brightness from 200 to 800,
the image is collected fom openart and processed which crope to simulate real situation process code can get form enhance_img branch
| model name | qvga_200 | qvga_400 | qvga_600 | qvga_800 |
|---|---|---|---|---|
| mobilenetv1_128 | 23.483% | 27.461% | 21.429% | 44.083% |
| djl_tupian_128 | 7.812% | 4.145% | 2.551% | 2.663% |