Res2Net

Introduction

We propose a novel building block for CNNs, namely Res2Net, by constructing hierarchical residual-like connections within one single residual block. The Res2Net represents multi-scale features at a granular level and increases the range of receptive fields for each network layer. The proposed Res2Net block can be plugged into the state-of-the-art backbone CNN models, e.g. , ResNet, ResNeXt, BigLittleNet, and DLA. We evaluate the Res2Net block on all these models and demonstrate consistent performance gains over baseline models.

Benchmark

Model Context Top-1 (%) Top-5 (%) Params (M) Train T. Infer T. Download Config Log
Res2Net50 D910x8-G 79.35 94.64 25.76 246s/epoch 28.5ms/step model cfg log
Res2Net101 D910x8-G 79.56 94.70 45.33 467s/epoch 46.0ms/step model cfg log
Res2Net50 D910x8-G 80.32 95.09 25.77 250s/epoch 29.6ms/step model cfg log
Res2Net101-v1b D910x8-G 81.26 95.41 45.35 435s/epoch 42.4ms/step model cfg log

Notes

  • All models are trained on ImageNet-1K training set and the top-1 accuracy is reported on the validatoin set.

  • Context: GPU_TYPE x pieces - G/F, G - graph mode, F - pynative mode with ms function.

Examples

Train

  • The yaml config files that yield competitive results on ImageNet for different models are listed in the configs folder. To trigger training using preset yaml config.

    mpirun -n 8 python train.py --config configs/res2net/res2net_50_gpu.yaml --data_dir /path/to/imagenet

Detailed adjustable parameters and their default value can be seen in config.py.

Eval

  • To validate the model, you can use validate.py. Here is an example for res2net50 to verify the accuracy of your training.

    python validate.py --config configs/res2net/res2net_50_gpu.yaml --data_dir /path/to/imagenet --ckpt_path /path/to/res2net50.ckpt