dpn

Dual Path Networks (DPN)

Dual Path Networks

Introduction

Figure 1 shows the model architecture of ResNet, DenseNet and Dual Path Networks. By combining the feature reusage of ResNet and new feature introduction of DenseNet, DPN could enjoy both benefits so that it could share common features and maintain the flexibility to explore new features. As a result, DPN could achieve better performance with fewer computation cost compared with ResNet and DenseNet on ImageNet-1K dataset.[1]

Figure 1. Architecture of DPN [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/dpn/dpn92_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/dpn/dpn92_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/dpn/dpn92_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.

coming soon

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
dpn92	37.79	8	32	224x224	O2	293s	78.22	3272.82	79.46	94.49	yaml	weights

Notes

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

References

[1] Chen Y, Li J, Xiao H, et al. Dual path networks[J]. Advances in neural information processing systems, 2017, 30.