MobileNetV1

Introduction

Compared with the traditional convolutional neural network, MobileNetV1’s parameters and the amount of computation are greatly reduced on the premise that the accuracy rate is slightly reduced. (Compared to VGG16, the accuracy rate is reduced by 0.9%, but the model parameters are only 1/32 of VGG). The model is based on a streamlined architecture that uses depthwise separable convolutions to build lightweight deep neural networks. At the same time, two simple global hyperparameters are introduced, which can effectively trade off latency and accuracy.

Benchmark

Pynative Pynative Graph Graph
Model Top-1 (%) Top-5 (%) train (s/epoch) Infer (ms) train(s/epoch) Infer (ms) Download Config
GPU MobileNet_v1_100 71.95 90.41 model config
Ascend MobileNet_v1_100 71.83 90.26
GPU MobileNet_v1_075 70.84 89.63 model config
Ascend MobileNet_v1_075 70.66 89.49
GPU MobileNet_v1_050 66.37 86.71 model config
Ascend MobileNet_v1_050 66.39 86.85
GPU MobileNet_v1_025 54.58 78.27 model config
Ascend MobileNet_v1_025 54.64 78.29

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.

    export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7
mpirun --allow-run-as-root -n 8 python train.py -c configs/mobilenetv1/mobilenetv1_075_gpu.yaml

  • Here is the example for finetuning a pretrained MobileNetV1 on CIFAR10 dataset using Adam optimizer.

    python train.py --model=mobilenet_v1_075_224 --pretrained --opt=adam --lr=0.001 dataset=cifar10 --num_classes=10 --dataset_download

    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 to verify the accuracy of pretrained weights.

    python validate.py --model=mobilenet_v1_075_224 --dataset=imagenet --val_split=val --pretrained

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

    python validate.py --model=mobilenet_v1_075_224 --dataset=imagenet --val_split=val --ckpt_path='./ckpt/mobilenet_v1_075_224-best.ckpt'