mindspore/model_zoo/official/cv/mobilenetv2

Contents

• MobileNetV2 Description
• Model Architecture
• Dataset
• Features
  • Mixed Precision
• Environment Requirements
• Script Description
  • Script and Sample Code
    • Training Process
    • Evaluation Process
      • Evaluation
• Model Description
  • Performance
    • Training Performance
    • Inference Performance
• Description of Random Situation
• ModelZoo Homepage

MobileNetV2 Description

MobileNetV2 is tuned to mobile phone CPUs through a combination of hardware- aware network architecture search (NAS) complemented by the NetAdapt algorithm and then subsequently improved through novel architecture advances.Nov 20, 2019.

Paper Howard, Andrew, Mark Sandler, Grace Chu, Liang-Chieh Chen, Bo Chen, Mingxing Tan, Weijun Wang et al. "Searching for MobileNetV2." In Proceedings of the IEEE International Conference on Computer Vision, pp. 1314-1324. 2019.

Model architecture

The overall network architecture of MobileNetV2 is show below:

Link

Dataset

Dataset used: imagenet

• Dataset size: ~125G, 1.2W colorful images in 1000 classes
  • Train: 120G, 1.2W images
  • Test: 5G, 50000 images
• Data format: RGB images.
  • Note: Data will be processed in src/dataset.py

Features

Mixed Precision(Ascend)

The mixed precision training method accelerates the deep learning neural network training process by using both the single-precision and half-precision data formats, and maintains the network precision achieved by the single-precision training at the same time. Mixed precision training can accelerate the computation process, reduce memory usage, and enable a larger model or batch size to be trained on specific hardware. For FP16 operators, if the input data type is FP32, the backend of MindSpore will automatically handle it with reduced precision. Users could check the reduced-precision operators by enabling INFO log and then searching ‘reduce precision’.

Environment Requirements

• Hardware（Ascend/GPU/CPU）
  • Prepare hardware environment with Ascend、GPU or CPU processor. If you want to try Ascend, please send the application form to ascend@huawei.com. Once approved, you can get the resources.
• Framework
  • MindSpore
• For more information, please check the resources below：
  • MindSpore tutorials
  • MindSpore API

Script description

Script and sample code

├── MobileNetV2
  ├── Readme.md     # descriptions about MobileNetV2
  ├── scripts
  │   ├──run_train.sh   # shell script for train, fine_tune or incremental  learn with CPU, GPU or Ascend
  │   ├──run_eval.sh    # shell script for evaluation with CPU, GPU or Ascend
  ├── src
  │   ├──args.py        # parse args
  │   ├──config.py      # parameter configuration
  │   ├──dataset.py     # creating dataset
  │   ├──launch.py      # start python script
  │   ├──lr_generator.py     # learning rate config
  │   ├──mobilenetV2.py      # MobileNetV2 architecture
  │   ├──models.py      # contain define_net and Loss, Monitor
  │   ├──utils.py       # utils to load ckpt_file for fine tune or incremental learn
  ├── train.py      # training script
  ├── eval.py       # evaluation script

Training process

Usage

You can start training using python or shell scripts. The usage of shell scripts as follows:

• Ascend: sh run_train.sh Ascend [DEVICE_NUM] [VISIABLE_DEVICES(0,1,2,3,4,5,6,7)] [RANK_TABLE_FILE] [DATASET_PATH] [TRAIN_METHOD] [CKPT_PATH]
• GPU: sh run_trian.sh GPU [DEVICE_NUM] [VISIABLE_DEVICES(0,1,2,3,4,5,6,7)] [DATASET_PATH] [TRAIN_METHOD] [CKPT_PATH]
• CPU: sh run_trian.sh CPU [DATASET_PATH] [TRAIN_METHOD] [CKPT_PATH]

Launch

# training example
  python:
      Ascend: python train.py --dataset_path ~/imagenet/train/ --platform Ascend --train_method train
      GPU: python train.py --dataset_path ~/imagenet/train/ --platform GPU --train_method train
      CPU: python train.py --dataset_path ~/imagenet/train/ --platform CPU --train_method train

  shell:
      Ascend: sh run_train.sh Ascend 8 0,1,2,3,4,5,6,7 hccl_config.json ~/imagenet/train/  train
      GPU: sh run_train.sh GPU 8 0,1,2,3,4,5,6,7 ~/imagenet/train/ train
      CPU: sh run_train.sh CPU ~/imagenet/train/ train

# fine tune example
  python:
      Ascend: python train.py --dataset_path ~/imagenet/train/ --platform Ascend --train_method fine_tune ./pretrain_checkpoint/mobilenetv2_199.ckpt
      GPU: python train.py --dataset_path ~/imagenet/train/ --platform GPU --train_method fine_tune ./pretrain_checkpoint/mobilenetv2_199.ckpt
      CPU: python train.py --dataset_path ~/imagenet/train/ --platform CPU --train_method fine_tune ./pretrain_checkpoint/mobilenetv2_199.ckpt

  shell:
      Ascend: sh run_train.sh Ascend 8 0,1,2,3,4,5,6,7 hccl_config.json ~/imagenet/train/  fine_tune ./pretrain_checkpoint/mobilenetv2_199.ckpt
      GPU: sh run_train.sh GPU 8 0,1,2,3,4,5,6,7 ~/imagenet/train/ fine_tune ./pretrain_checkpoint/mobilenetv2_199.ckpt
      CPU: sh run_train.sh CPU ~/imagenet/train/ fine_tune ./pretrain_checkpoint/mobilenetv2_199.ckpt

# incremental learn example
  python:
      Ascend: python train.py --dataset_path ~/imagenet/train/ --platform Ascend --train_method incremental_learn ./pretrain_checkpoint/mobilenetv2_199.ckpt
      GPU: python train.py --dataset_path ~/imagenet/train/ --platform GPU --train_method incremental_learn ./pretrain_checkpoint/mobilenetv2_199.ckpt
      CPU: python train.py --dataset_path ~/imagenet/train/ --platform CPU --train_method incremental_learn ./pretrain_checkpoint/mobilenetv2_199.ckpt

  shell:
      Ascend: sh run_train.sh Ascend 8 0,1,2,3,4,5,6,7 hccl_config.json ~/imagenet/train/  incremental_learn ./pretrain_checkpoint/mobilenetv2_199.ckpt
      GPU: sh run_train.sh GPU 8 0,1,2,3,4,5,6,7 ~/imagenet/train/ incremental_learn ./pretrain_checkpoint/mobilenetv2_199.ckpt
      CPU: sh run_train.sh CPU ~/imagenet/train/ incremental_learn ./pretrain_checkpoint/mobilenetv2_199.ckpt

Result

Training result will be stored in the example path. Checkpoints will be stored at . /checkpoint by default, and training log will be redirected to ./train/train.log like followings.

epoch: [  0/200], step:[  624/  625], loss:[5.258/5.258], time:[140412.236], lr:[0.100]
epoch time: 140522.500, per step time: 224.836, avg loss: 5.258
epoch: [  1/200], step:[  624/  625], loss:[3.917/3.917], time:[138221.250], lr:[0.200]
epoch time: 138331.250, per step time: 221.330, avg loss: 3.917

Eval process

Usage

You can start training using python or shell scripts. The usage of shell scripts as follows:

• Ascend: sh run_infer.sh Ascend [DATASET_PATH] [CHECKPOINT_PATH] [HEAD_CKPT_PATH]
• GPU: sh run_infer.sh GPU [DATASET_PATH] [CHECKPOINT_PATH] [HEAD_CKPT_PATH]
• CPU: sh run_infer.sh CPU [DATASET_PATH] [BACKBONE_CKPT_PATH] [HEAD_CKPT_PATH]

Launch

# infer example
  python:
      Ascend: python eval.py --dataset_path ~/imagenet/val/ --pretrain_ckpt ~/train/mobilenet-200_625.ckpt --platform Ascend --head_ckpt ./checkpoint/mobilenetv2_199.ckpt
      GPU: python eval.py --dataset_path ~/imagenet/val/ --pretrain_ckpt ~/train/mobilenet-200_625.ckpt --platform GPU --head_ckpt ./checkpoint/mobilenetv2_199.ckpt
      CPU: python eval.py --dataset_path ~/imagenet/val/ --pretrain_ckpt ~/train/mobilenet-200_625.ckpt --platform CPU --head_ckpt ./checkpoint/mobilenetv2_199.ckpt

  shell:
      Ascend: sh run_infer.sh Ascend ~/imagenet/val/ ~/train/mobilenet-200_625.ckpt ./checkpoint/mobilenetv2_199.ckpt
      GPU: sh run_infer.sh GPU ~/imagenet/val/ ~/train/mobilenet-200_625.ckpt ./checkpoint/mobilenetv2_199.ckpt
      CPU: sh run_infer.sh GPU ~/imagenet/val/ ~/train/mobilenet-200_625.ckpt ./checkpoint/mobilenetv2_199.ckpt

checkpoint can be produced in training process.

Result

Inference result will be stored in the example path, you can find result like the followings in val.log.

result: {'acc': 0.71976314102564111} ckpt=/path/to/checkpoint/mobilenet-200_625.ckpt

Model description

Performance

Training Performance

Parameters	MobilenetV2
Model Version		large
Resource	Ascend 910, cpu:2.60GHz 56cores, memory:314G	NV SMX2 V100-32G
uploaded Date	05/06/2020	05/06/2020
MindSpore Version	0.3.0	0.3.0
Dataset	ImageNet	ImageNet
Training Parameters	src/config.py	src/config.py
Optimizer	Momentum	Momentum
Loss Function	SoftmaxCrossEntropy	SoftmaxCrossEntropy
outputs
Loss		1.913
Accuracy		ACC1[77.09%] ACC5[92.57%]
Total time
Params (M)
Checkpoint for Fine tuning
Model for inference

Inference Performance

Parameters
Model Version	V1
Resource	Ascend 910	NV SMX2 V100-32G	Ascend 310
uploaded Date	05/06/2020	05/22/2020
MindSpore Version	0.2.0	0.2.0	0.2.0
Dataset	ImageNet, 1.2W	ImageNet, 1.2W	ImageNet, 1.2W
batch_size		130(8P)
outputs
Accuracy		ACC1[72.07%] ACC5[90.90%]
Speed
Total time
Model for inference

Description of Random Situation

In dataset.py, we set the seed inside “create_dataset" function. We also use random seed in train.py.

ModelZoo Homepage

Please check the official homepage.