From 0f0a7923424ac1c21844d32490f5306e0da468e6 Mon Sep 17 00:00:00 2001 From: iamhankai Date: Fri, 29 Nov 2019 20:25:50 +0800 Subject: [PATCH] update readme --- readme.md | 15 +++++++++------ 1 file changed, 9 insertions(+), 6 deletions(-) diff --git a/readme.md b/readme.md index c25fbec..639d85d 100755 --- a/readme.md +++ b/readme.md @@ -1,6 +1,6 @@ # GhostNet -GhostNet: More Features from Cheap Operations +GhostNet: More Features from Cheap Operations [[arXiv]](https://arxiv.org/abs/1911.11907) By Kai Han, Yunhe Wang, Qi Tian, Jianyuan Guo, Chunjing Xu, Chang Xu. @@ -12,22 +12,25 @@ By Kai Han, Yunhe Wang, Qi Tian, Jianyuan Guo, Chunjing Xu, Chang Xu. - **Performance** +We beat other SOTA lightweight CNNs such as **MobileNetV3** and **FBNet**. +
## Implementation -The code provides the inference code and pretrained model of GhostNet on ImageNet. +The code provides the TensorFlow code and pretrained model of GhostNet on ImageNet. `myconv2d.py` implemented `GhostModule` and `ghost_net.py` implemented `GhostNet`. +### Requirements +The code was verified on Python3.6, TensorFlow-1.13.1, Tensorpack-0.9.7. Not sure on other version. + ### Usage Run `python test-ghostnet.py --eval --data_dir=/path/to/imagenet/dir/ --load=./models/ghostnet_checkpoint` to evaluate on `val` set. -You'll get the accuracy: top-1 error=`0.26066`, top-5 error=`0.08614`. -### Requirements -The code was verified on Python3.6, TF-1.13.1, Tensorpack-0.9.7. Not sure on other version. +You'll get the accuracy: top-1 error=`0.26066`, top-5 error=`0.08614` with only `141M` Flops (or say MAdds). ### Data Preparation ImageNet data dir should have the following structure, and `val` and `caffe_ilsvrc12` subdirs are essential: @@ -50,7 +53,7 @@ caffe_ilsvrc12 data can be downloaded from http://dl.caffe.berkeleyvision.org/ca @article{ghostnet, title={GhostNet: More Features from Cheap Operations}, author={Han, Kai and Wang, Yunhe and Tian, Qi and Guo, Jianyuan and Xu, Chunjing and Xu, Chang}, - journal={arXiv preprint}, + journal={arXiv}, year={2019} } ```