search

mit-han-lab / mcunet

[NeurIPS 2020] MCUNet: Tiny Deep Learning on IoT Devices; [NeurIPS 2021] MCUNetV2: Memory-Efficient Patch-based Inference for Tiny Deep Learning
https://mcunet.mit.edu
MIT License
460 stars 82 forks source link

validation dataset clarification for evaluation #2

Closed ghost closed 2 years ago

ghost commented 2 years ago

I am confused about the val dataset, is it image net-1000 classes or tiny-image net-200, also what is the input size of images trained is it 64x64 ?.

I tested the models and ran the script given in eval on github page, my dataset is tiny-imagenet-200 dataset, the accuracy shown is very low.

tonylins commented 2 years ago

Hi, the dataset is the official ImageNet-1k dataset (ILSVRC2012). Since the memory is quite limited for MCUs, we used a small input image resolution, which is designed by the Neural Architecture Search process.