PP-HGNet is a high-performance backbone network developed by the PaddleCV team that is more suitable for GPU devices. Compared with other SOTA models on GPU devices, this model has higher accuracy under the same latency. At the same latency, the model is 3.8 percentage points higher than the ResNet34-D model, 2.4 percentage points higher than the ResNet50-D model, and 4.7 percentage points higher than the ResNet50-D model after using the SSLD distillation strategy. At the same time, under the same accuracy, its latency is much smaller than the mainstream VisionTransformer model. We will release the technical report to arxiv recently, so stay tuned.
The author of PP-HGNet analyzes and summarizes the current GPU-friendly networks for GPU devices, and uses 3x3 standard convolutions as much as possible (the highest computational density). Here, VOVNet is used as the base model, and the improvement points that are mainly beneficial to GPU acceleration will be integrated. In the end, under the same latency of PP-HGNet, the accuracy greatly surpasses other backbones
The overall structure of the PP-HGNet backbone network is as follows:
Among them, PP-HGNet is composed of multiple HG-Blocks. The details of HG-Blocks are as follows:
The accuracy and latency of PP-HGNet are as follows:
| Model | Top-1 Acc(%) | Top-5 Acc(%) | Latency(ms) | Pre-trained model download address | Inference model download address |
|---|---|---|---|---|---|
| PPHGNet_tiny | 79.83 | 95.04 | 1.77 | download link | download link |
| PPHGNet_tiny_ssld | 81.95 | 96.12 | 1.77 | download link | download link |
| PPHGNet_small | 81.51 | 95.82 | 2.52 | download link | download link |
| PPHGNet_small_ssld | 83.82 | 96.81 | 2.52 | download link | download link |
| PPHGNet_base_ssld | 85.00 | 97.35 | 5.97 | download link | download link |
Note:
-
_ssldrepresents the model after usingSSLD distillation. For details aboutSSLD distillation, see SSLD distillation.
-
- More metrics and weights of PP-HGNet, so stay tuned.
The comparison between PP-HGNet and other models is as follows. The test machine is NVIDIA® Tesla® V100, the TensorRT engine is turned on, and the precision type is FP32. Under the same latency, the accuracy of PP-HGNet surpasses other SOTA CNN models, and in comparison with the SwinTransformer model, it is more than 2 times faster with higher accuracy.
| Model | Top-1 Acc(%) | Top-5 Acc(%) | Latency(ms) |
|---|---|---|---|
| ResNet34 | 74.57 | 92.14 | 1.97 |
| ResNet34_vd | 75.98 | 92.98 | 2.00 |
| EfficientNetB0 | 77.38 | 93.31 | 1.96 |
| PPHGNet_tiny | 79.83 | 95.04 | 1.77 |
| PPHGNet_tiny_ssld | 81.95 | 96.12 | 1.77 |
| ResNet50 | 76.50 | 93.00 | 2.54 |
| ResNet50_vd | 79.12 | 94.44 | 2.60 |
| ResNet50_rsb | 80.40 | 2.54 | |
| EfficientNetB1 | 79.15 | 94.41 | 2.88 |
| SwinTransformer_tiny | 81.2 | 95.5 | 6.59 |
| PPHGNet_small | 81.51 | 95.82 | 2.52 |
| PPHGNet_small_ssld | 83.82 | 96.81 | 2.52 |
| Res2Net200_vd_26w_4s_ssld | 85.13 | 97.42 | 11.45 |
| ResNeXt101_32x48d_wsl | 85.37 | 97.69 | 55.07 |
| SwinTransformer_base | 85.2 | 97.5 | 13.53 |
| PPHGNet_base_ssld | 85.00 | 97.35 | 5.97 |
- Run the following command to install if CUDA9 or CUDA10 is available.
python3 -m pip install paddlepaddle-gpu -i https://mirror.baidu.com/pypi/simple- Run the following command to install if GPU device is unavailable.
python3 -m pip install paddlepaddle -i https://mirror.baidu.com/pypi/simplePlease refer to PaddlePaddle Installation for more information about installation, for examples other versions.
The command of PaddleClas installation as bellow:
pip3 install paddleclas
- Prediction with CLI
paddleclas --model_name=PPHGNet_small --infer_imgs="docs/images/inference_deployment/whl_demo.jpg"Results:
>>> result
class_ids: [8, 7, 86, 82, 81], scores: [0.71479, 0.08682, 0.00806, 0.0023, 0.00121], label_names: ['hen', 'cock', 'partridge', 'ruffed grouse, partridge, Bonasa umbellus', 'ptarmigan'], filename: docs/images/inference_deployment/whl_demo.jpg
Predict complete!
Note: When replacing other scale models of PPHGNet, just replace model_name. For example, when changing the model at this time to PPHGNet_tiny, you only need to change --model_name=PPHGNet_small to --model_name=PPHGNet_tiny.
- Prediction in Python
from paddleclas import PaddleClas
clas = PaddleClas(model_name='PPHGNet_small')
infer_imgs = 'docs/images/inference_deployment/whl_demo.jpg'
result = clas.predict(infer_imgs)
print(next(result))The result of demo above:
>>> result
[{'class_ids': [8, 7, 86, 82, 81], 'scores': [0.77132, 0.05122, 0.00755, 0.00199, 0.00115], 'label_names': ['hen', 'cock', 'partridge', 'ruffed grouse, partridge, Bonasa umbellus', 'ptarmigan'], 'filename': 'docs/images/inference_deployment/whl_demo.jpg'}]
Note: The result returned by model.predict() is a generator, so you need to use the next() function to call it or for loop to loop it. And it will predict with batch_size size batch and return the prediction results when called. The default batch_size is 1, and you also specify the batch_size when instantiating, such as model = paddleclas.PaddleClas(model_name="PPHGNet_small", batch_size=2).
Please refer to Installation to get the description about installation.
Please prepare ImageNet-1k data at ImageNet official website.
Enter the PaddleClas/ directory:
cd path_to_PaddleClas
Enter the dataset/ directory, name the downloaded data ILSVRC2012 , and the ILSVRC2012 directory has the following data:
├── train
│ ├── n01440764
│ │ ├── n01440764_10026.JPEG
│ │ ├── n01440764_10027.JPEG
├── train_list.txt
...
├── val
│ ├── ILSVRC2012_val_00000001.JPEG
│ ├── ILSVRC2012_val_00000002.JPEG
├── val_list.txt
where train/ and val/ are the training set and validation set, respectively. train_list.txt and val_list.txt are the label files for the training set and validation set, respectively.
Note:
- About the contents format of
train_list.txtandval_list.txt, please refer to Description about Classification Dataset in PaddleClas.
The PPHGNet_small training configuration is provided in ppcls/configs/ImageNet/PPHGNet/PPHGNet_small.yaml, which can be started with the following script:
export CUDA_VISIBLE_DEVICES=0,1,2,3
python3 -m paddle.distributed.launch \
--gpus="0,1,2,3" \
tools/train.py \
-c ppcls/configs/ImageNet/PPHGNet/PPHGNet_small.yamlNote:
- The current model with the best accuracy will be saved in
output/PPHGNet_small/best_model.pdparams
If you are not training an ImageNet task, you need to change the configuration file and training method, such as reducing the learning rate, reducing the number of epochs, etc.
After training, you can use the following commands to evaluate the model.
python3 tools/eval.py \
-c ppcls/configs/ImageNet/PPHGNet/PPHGNet_small.yaml \
-o Global.pretrained_model=output/PPHGNet_small/best_modelAmong the above command, the argument -o Global.pretrained_model="output/PPHGNet_small/best_model" specify the path of the best model weight file. You can specify other path if needed.
After the model training is completed, the pre-trained model obtained from the training can be loaded for model prediction. A complete example is provided in the tools/infer.py of the model library, and the model prediction can be done by simply executing the following command:
python3 tools/infer.py \
-c ppcls/configs/ImageNet/PPHGNet/PPHGNet_small.yaml \
-o Global.pretrained_model=output/PPHGNet_small/best_modelThe results:
[{'class_ids': [8, 7, 86, 82, 81], 'scores': [0.71479, 0.08682, 0.00806, 0.0023, 0.00121], 'file_name': 'docs/images/inference_deployment/whl_demo.jpg', 'label_names': ['hen', 'cock', 'partridge', 'ruffed grouse, partridge, Bonasa umbellus', 'ptarmigan']}]
Note:
-
Among the above command, argument
-o Global.pretrained_model="output/PPHGNet_small/best_model"specify the path of the best model weight file. You can specify other path if needed. -
The default test image is
docs/images/inference_deployment/whl_demo.jpg，And you can test other image, only need to specify the argument-o Infer.infer_imgs=path_to_test_image. -
The default output is the value of Top-5. If you want to output the value of Top-k, you can specify
-o Infer.PostProcess.topk=k, wherekis the value you specify. -
The default label mapping is based on the ImageNet dataset. If you change the dataset, you need to re-specify
Infer.PostProcess.class_id_map_file. For the method of making the mapping file, please refer toppcls/utils/imagenet1k_label_list.txt
Paddle Inference is the original Inference Library of the PaddlePaddle, provides high-performance inference for server deployment. And compared with directly based on the pretrained model, Paddle Inference can use tools to accelerate prediction, so as to achieve better inference performance. Please refer to Paddle Inference for more information.
Paddle Inference need Paddle Inference Model to predict. Two process provided to get Paddle Inference Model. If want to use the provided by PaddleClas, you can download directly, click Downloading Inference Model.
The command about exporting Paddle Inference Model is as follow:
python3 tools/export_model.py \
-c ppcls/configs/ImageNet/PPHGNet/PPHGNet_small.yaml \
-o Global.pretrained_model=output/PPHGNet_small/best_model \
-o Global.save_inference_dir=deploy/models/PPHGNet_small_inferAfter running above command, the inference model files would be saved in deploy/models/PPHGNet_small_infer, as shown below:
├── PPHGNet_small_infer
│ ├── inference.pdiparams
│ ├── inference.pdiparams.info
│ └── inference.pdmodel
You can also download directly.
cd deploy/models
# download the inference model and decompression
wget https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPHGNet_small_infer.tar && tar -xf PPHGNet_small_infer.tar
After decompression, the directory models should be shown below.
├── PPHGNet_small_infer
│ ├── inference.pdiparams
│ ├── inference.pdiparams.info
│ └── inference.pdmodel
Return the directory deploy:
cd ../
Run the following command to classify whether there are humans in the image ./images/ImageNet/ILSVRC2012_val_00000010.jpeg.
# Use the following command to predict with GPU.
python3 python/predict_cls.py -c configs/inference_cls.yaml -o Global.inference_model_dir=models/PPHGNet_small_infer
# Use the following command to predict with CPU.
python3 python/predict_cls.py -c configs/inference_cls.yaml -o Global.inference_model_dir=models/PPHGNet_small_infer -o Global.use_gpu=FalseThe prediction results:
ILSVRC2012_val_00000010.jpeg: class id(s): [332, 153, 283, 338, 204], score(s): [0.50, 0.05, 0.02, 0.01, 0.01], label_name(s): ['Angora, Angora rabbit', 'Maltese dog, Maltese terrier, Maltese', 'Persian cat', 'guinea pig, Cavia cobaya', 'Lhasa, Lhasa apso']
If you want to predict images in directory, please specify the argument Global.infer_imgs as directory path by -o Global.infer_imgs. The command is as follow.
# Use the following command to predict with GPU. If want to replace with CPU, you can add argument -o Global.use_gpu=False
python3 python/predict_cls.py -c configs/inference_cls.yaml -o Global.inference_model_dir=models/PPHGNet_small_infer -o Global.infer_imgs=images/ImageNet/终端中会输出该文件夹内所有图像的分类结果，如下所示。
ILSVRC2012_val_00000010.jpeg: class id(s): [332, 153, 283, 338, 204], score(s): [0.50, 0.05, 0.02, 0.01, 0.01], label_name(s): ['Angora, Angora rabbit', 'Maltese dog, Maltese terrier, Maltese', 'Persian cat', 'guinea pig, Cavia cobaya', 'Lhasa, Lhasa apso']
ILSVRC2012_val_00010010.jpeg: class id(s): [626, 622, 531, 487, 633], score(s): [0.68, 0.02, 0.02, 0.02, 0.02], label_name(s): ['lighter, light, igniter, ignitor', 'lens cap, lens cover', 'digital watch', 'cellular telephone, cellular phone, cellphone, cell, mobile phone', "loupe, jeweler's loupe"]
ILSVRC2012_val_00020010.jpeg: class id(s): [178, 211, 171, 246, 741], score(s): [0.82, 0.00, 0.00, 0.00, 0.00], label_name(s): ['Weimaraner', 'vizsla, Hungarian pointer', 'Italian greyhound', 'Great Dane', 'prayer rug, prayer mat']
ILSVRC2012_val_00030010.jpeg: class id(s): [80, 83, 136, 23, 93], score(s): [0.84, 0.00, 0.00, 0.00, 0.00], label_name(s): ['black grouse', 'prairie chicken, prairie grouse, prairie fowl', 'European gallinule, Porphyrio porphyrio', 'vulture', 'hornbill']
PaddleClas provides an example about how to deploy with C++. Please refer to Deployment with C++.
Paddle Serving is a flexible, high-performance carrier for machine learning models, and supports different protocol, such as RESTful, gRPC, bRPC and so on, which provides different deployment solutions for a variety of heterogeneous hardware and operating system environments. Please refer Paddle Serving for more information.
PaddleClas provides an example about how to deploy as service by Paddle Serving. Please refer to Paddle Serving Deployment.
Paddle-Lite is an open source deep learning framework that designed to make easy to perform inference on mobile, embeded, and IoT devices. Please refer to Paddle-Lite for more information.
PaddleClas provides an example of how to deploy on mobile by Paddle-Lite. Please refer to Paddle-Lite deployment.
Paddle2ONNX support convert Paddle Inference model to ONNX model. And you can deploy with ONNX model on different inference engine, such as TensorRT, OpenVINO, MNN/TNN, NCNN and so on. About Paddle2ONNX details, please refer to Paddle2ONNX.
PaddleClas provides an example of how to convert Paddle Inference model to ONNX model by paddle2onnx toolkit and predict by ONNX model. You can refer to paddle2onnx for deployment details.