vgg

VGGNet

Very Deep Convolutional Networks for Large-Scale Image Recognition

Introduction

Figure 1 shows the model architecture of VGGNet. VGGNet is a key milestone on image classification task. It expands the model to 16-19 layers for the first time. The key motivation of this model is that it shows usage of 3x3 kernels is efficient and by adding 3x3 kernels, it could have the same effect as 5x5 or 7x7 kernels. VGGNet could achieve better model performance compared with previous methods such as GoogleLeNet and AlexNet on ImageNet-1K dataset.[1]

Figure 1. Architecture of VGG [1]

Requirements

mindspore	ascend driver	firmware	cann toolkit/kernel
2.3.1	24.1.RC2	7.3.0.1.231	8.0.RC2.beta1

Quick Start

Preparation

Installation

Please refer to the installation instruction in MindCV.

Dataset Preparation

Please download the ImageNet-1K dataset for model training and validation.

Training

• Distributed Training

It is easy to reproduce the reported results with the pre-defined training recipe. For distributed training on multiple Ascend 910 devices, please run

# distrubted training on multiple GPU/Ascend devices
msrun --bind_core=True --worker_num 8 python train.py --config configs/vgg/vgg16_ascend.yaml --data_dir /path/to/imagenet

For detailed illustration of all hyper-parameters, please refer to config.py.

Note: As the global batch size (batch_size x num_devices) is an important hyper-parameter, it is recommended to keep the global batch size unchanged for reproduction or adjust the learning rate linearly to a new global batch size.

• Standalone Training

If you want to train or finetune the model on a smaller dataset without distributed training, please run:

# standalone training on single NPU device
python train.py --config configs/vgg/vgg16_ascend.yaml --data_dir /path/to/dataset --distribute False

Validation

To validate the accuracy of the trained model, you can use validate.py and parse the checkpoint path with --ckpt_path.

python validate.py -c configs/vgg/vgg16_ascend.yaml --data_dir /path/to/imagenet --ckpt_path /path/to/ckpt

Performance

Our reproduced model performance on ImageNet-1K is reported as follows.

Experiments are tested on ascend 910* with mindspore 2.3.1 graph mode.

model name	params(M)	cards	batch size	resolution	jit level	graph compile	ms/step	img/s	acc@top1	acc@top5	recipe	weight
vgg13	133.04	8	32	224x224	O2	41s	30.52	8387.94	72.81	91.02	yaml	weights
vgg19	143.66	8	32	224x224	O2	53s	39.17	6535.61	75.24	92.55	yaml	weights

Experiments are tested on ascend 910 with mindspore 2.3.1 graph mode.

model name	params(M)	cards	batch size	resolution	jit level	graph compile	ms/step	img/s	acc@top1	acc@top5	recipe	weight
vgg13	133.04	8	32	224x224	O2	23s	55.20	4637.68	72.87	91.02	yaml	weights
vgg19	143.66	8	32	224x224	O2	22s	67.42	3797.09	75.21	92.56	yaml	weights

Notes

• top-1 and top-5: Accuracy reported on the validation set of ImageNet-1K.

References

[1] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]. arXiv preprint arXiv:1409.1556, 2014.