This is the PyTorch implement of MobileNet V2
Branch: master
Clone or download
hzweikaifeng update
Latest commit e43c9b9 Jan 30, 2018
Type Name Latest commit message Commit time
Failed to load latest commit information.
ImageNet/ILSVRC2012_devkit_t12/data original commit Jan 23, 2018 update readme Jan 26, 2018 fix validation dataset label Jan 26, 2018 update Jan 30, 2018

This is the PyTorch implement of MobileNet V2 (train on ImageNet dataset)

Paper: Inverted Residuals and Linear Bottlenecks: Mobile Networks for Classification, Detection and Segment


Prepare data

This code takes ImageNet dataset as example. You can download ImageNet dataset and put them as follows. I only provide ILSVRC2012_dev_kit_t12 due to the restriction of memory, in other words, you need download ILSVRC2012_img_train and ILSVRC2012_img_val.

├── # train script
├── # network of MobileNetV2
├── # ImageNet dataset read script
├── ImageData # train and validation data
	├── ILSVRC2012_img_train
		├── n01440764
		├──    ...
		├── n15075141
	├── ILSVRC2012_img_val
	├── ILSVRC2012_dev_kit_t12
		├── data
			├── ILSVRC2012_validation_ground_truth.txt
			├── meta.mat # the map between train file name and label


  • If you want to train from scratch, you can run as follows:
python --batch-size 256 --gpus 0,1,2,3
  • If you want to train from one checkpoint, you can run as follows(for example train from epoch_4.pth.tar, the --start-epoch parameter is corresponding to the epoch of the checkpoint):
python --batch-size 256 --gpus 0,1,2,3 --resume output/epoch_4.pth.tar --start-epoch 4