mindspore/model_zoo/research/cv/ghostnet_quant/Readme.md

Contents

• GhostNet Description
• Quantization Description
• Model Architecture
• Dataset
• 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

GhostNet Description

The GhostNet architecture is based on an Ghost module structure which generate more features from cheap operations. Based on a set of intrinsic feature maps, a series of cheap operations are applied to generate many ghost feature maps that could fully reveal information underlying intrinsic features.

Paper: Kai Han, Yunhe Wang, Qi Tian, Jianyuan Guo, Chunjing Xu, Chang Xu. GhostNet: More Features from Cheap Operations. CVPR 2020.

Quantization Description

Quantization refers to techniques for performing computations and storing tensors at lower bitwidths than floating point precision. For 8bit quantization, we quantize the weights into [-128,127] and the activations into [0,255]. We finetune the model a few epochs after post-quantization to achieve better performance.

Model architecture

The overall network architecture of GhostNet is show below:

Link

Dataset

Dataset used: Oxford-IIIT Pet

• Dataset size: 7049 colorful images in 1000 classes
  • Train: 3680 images
  • Test: 3369 images
• Data format: RGB images.
  • Note: Data will be processed in src/dataset.py

Environment Requirements

• Hardware（Ascend/GPU）
  • Prepare hardware environment with Ascend or GPU 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 Python API

Script description

Script and sample code

├── GhostNet
  ├── Readme.md     # descriptions about GhostNet   # shell script for evaluation with CPU, GPU or Ascend
  ├── src
  │   ├──config.py      # parameter configuration
  │   ├──dataset.py     # creating dataset
  │   ├──launch.py      # start python script
  │   ├──lr_generator.py     # learning rate config
  │   ├──ghostnet.py      # GhostNet architecture
  │   ├──quant.py      # GhostNet quantization
  ├── eval.py       # evaluation script
  ├── mindspore_hub_conf.py       # export model for hub

Training process

To Be Done

Eval process

Usage

After installing MindSpore via the official website, you can start evaluation as follows:

Launch

# infer example
  
  Ascend: python eval.py --dataset_path ~/Pets/test.mindrecord --platform Ascend --checkpoint_path [CHECKPOINT_PATH]
  GPU: python eval.py --dataset_path ~/Pets/test.mindrecord --platform GPU --checkpoint_path [CHECKPOINT_PATH]

checkpoint can be produced in training process.

Result

result: {'acc': 0.825} ckpt= ./ghostnet_1x_pets_int8.ckpt

Model Description

Performance

Evaluation Performance

GhostNet on ImageNet2012

Parameters
Model Version	GhostNet	GhostNet-int8
uploaded Date	09/08/2020 (month/day/year) ；	09/08/2020 (month/day/year)
MindSpore Version	0.6.0-alpha	0.6.0-alpha
Dataset	ImageNet2012	ImageNet2012
Parameters (M)	5.2	/
FLOPs (M)	142	/
Accuracy (Top1)	73.9	w/o finetune:72.2, w finetune:73.6

GhostNet on Oxford-IIIT Pet

Parameters
Model Version	GhostNet	GhostNet-int8
uploaded Date	09/08/2020 (month/day/year) ；	09/08/2020 (month/day/year)
MindSpore Version	0.6.0-alpha	0.6.0-alpha
Dataset	Oxford-IIIT Pet	Oxford-IIIT Pet
Parameters (M)	3.9	/
FLOPs (M)	140	/
Accuracy (Top1)	82.4	w/o finetune:81.66, w finetune:82.45

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.