-[1.1 Introduction to Knowledge Distillation](#1.1)
-[1.1.1 Response based distillation](#1.1.1)
-[1.1.2 Feature based distillation](#1.1.2)
-[1.1.3 Relation based distillation](#1.1.3)
-[1.2 Knowledge Distillation Algorithms Supported by PaddleClas](#1.2)
-[1.2.1 SSLD](#1.2.1)
-[1.2.2 DML](#1.2.2)
-[1.2.3 UDML](#1.2.3)
-[1.2.4 AFD](#1.2.4)
-[1.2.5 DKD](#1.2.5)
-[1.2.6 DIST](#1.2.6)
-[1.2.7 MGD](#1.2.7)
-[1.2.8 WSL](#1.2.8)
-[2. Usage](#2)
-[2.1 Environment Configuration](#2.1)
-[2.2 Data Preparation](#2.2)
-[2.3 Model Training](#2.3)
-[2.4 Model Evaluation](#2.4)
-[2.5 Model Prediction](#2.5)
-[2.6 Model Export & Inference](#2.6)
-[3. References](#3)
<aname="1"></a>
## 1. Introduction
<aname="1.1"></a>
### 1.1 Introduction to Knowledge Distillation
In recent years, deep neural network has been proved to be an effective method to solve problems in computer vision, natural language processing and other fields. By constructing an appropriate neural network and training it, the model performance will basically exceed traditional algorithms.
When data is sufficient, the model performance can be significantly improved by increasing the number of parameters through appropriate construction of the network, but this increases the model complexity. Large models are expensive to deploy in actual scenarios.
Redundancy exists in deep neural networks. At present, there are several methods to compress the model to reduce its parameter amount, e.g. pruning, quantization, knowledge distillation, etc. Knowledge distillation refers to the method that helps the training process of a smaller network (student) under the supervision of a larger network (teacher), so as to ensure the small model can obtain relatively large performance improvement and even obtain accuracy similar to large models without increasing parameters.
Knowledge distillation methods can be divided into three different categories: Response based distillation, Feature based distillation, Relation based distillation. Detailed introduction is as follows.
<aname='1.1.1'></a>
#### 1.1.1 Response based distillation
Knowledge distillation (KD) was first proposed by Hinton, who introduced KL divergence to the training loss function in addition to the cross entropy between the output logits and the ground truth labels. The accuracy of models trained with KD exceeds the accuracy of the same models trained only using ground truth loss. It should be noted that a larger teacher model needs to be trained first to guide the training process of student models.
PaddleClas proposed a simple yet effective SSLD algorithm [6], removing the dependence on ground truth labels. Combined with a large number of unlabeled data, the accuracy of pretrained models obtained from distillation on 18 models was improved by 3+%.
The aforementioned standard distillation method uses large models as teacher models to guide students to improve the performance. Later, Deep Mutual Learning (DML) distillation method [7] was proposed, i.e., two models with the same architecture learn from each other. Compared with KD and other knowledge distillation algorithms that rely on large teacher models, DML is independent of large teacher models. Such training process is simpler and more efficient.
<aname='1.1.2'></a>
#### 1.1.2 Feature based distillation
Heo et al porposed OverHaul of Feature Distillation [8], in which feature map distance between the teacher and the student is calculated as distillation loss. Features of the student are transformed to match the shape of the teacher's features so that the distance can be computed.
Knowledge distillation methods based on feature map distance can be combined with response-based knowledge distillation algorithms mentioned in `3.1`, i.e., the student's outputs and its middle feature maps are supervised simultaneously. For DML, such combination is even simpler, since the student's features can be aligned with the teacher's features without transformation. This method is used in PP-OCRv2, improving the accuracy of OCR models significantly.
<aname='1.1.3'></a>
#### 1.1.3 Relation based distillation
Papers in [1.1.1](#1.1.1) and [1.1.2](#1.1.2) mainly consider the outputs and middle feature maps of student and teacher. These distillation algorithms focus on individual outputs and do not consider relations between individuals.
Park et al proposed RKD [10], a distillation algorithm based on relations. In RKD, mutual relations of data examples is considered, and two loss functions are used, the distance-wise distillation loss and angle-wise distillation loss.
The algorithm proposed in this paper, Relational Knowledge Distillation (RKD), transfers the structured relations between the output results obtained from the teacher model to the student model. Unlike the previous algorithms, which only focus on individual output results, the RKD algorithm uses two loss functions: distance-wise distillation loss and angle-wise distillation loss. In the final distillation loss function, both KD loss and RKD loss are considered. The final accuracy is better than that obtained by KD loss distillation only.
<aname='1.2'></a>
### 1.2 Knowledge Distillation Algorithms in PaddleClas
<aname='1.2.1'></a>
#### 1.2.1 SSLD
##### 1.2.1.1 Introduction to SSLD
Paper:
> [Beyond Self-Supervision: A Simple Yet Effective Network Distillation Alternative to Improve Backbones](https://arxiv.org/abs/2103.05959)
>
> Cheng Cui, Ruoyu Guo, Yuning Du, Dongliang He, Fu Li, Zewu Wu, Qiwen Liu, Shilei Wen, Jizhou Huang, Xiaoguang Hu, Dianhai Yu, Errui Ding, Yanjun Ma
>
> arxiv, 2021
SSLD is a simple semi-supervised distillation method proposed by Baidu in 2021. By designing an improved JS divergence as the loss function and combining the data mining strategy based on ImageNet22k dataset, the accuracy of the 18 backbone network models was improved by more than 3% on average.
For more information about the principle, model zoo and usage of SSLD, please refer to: [Introduction to SSLD](ssld.md).
##### 1.2.1.2 Configuration of SSLD
The SSLD configuration is shown below. In the `Arch` field, you need to define both the student model and the teacher model. The teacher model has fixed parameters, and the pretrained parameters are loaded. In the `Loss` field, you need to define `DistillationDMLLoss` as the training loss.
```yaml
# model architecture
Arch:
name:"DistillationModel"# model name, here distillation model is used
class_num:&class_num1000# number of classes, 1000 for ImageNet1k
pretrained_list:# list of pretrained models, leave blank because specified in sub-models
freeze_params_list:# list of freezed params, networks of correspondent index are fixed if set to True
-True
-False
infer_model_name:"Student"# export Student sub-network when exporting model
models:# list of sub-networks
-Teacher:# teacher model
name:ResNet50_vd# model name
class_num:*class_num# number of classes
pretrained:True# pretrained model path, download official pretrained model if set to True
use_ssld:True# whether SSLD pretrained model is used (higher accuracy)
-Student:# student model
name:PPLCNet_x2_5# model name
class_num:*class_num# number of classes
pretrained:False# pretrained model path, can be bool or string. Set to False here. Student model does not load the pretrained model by default
# loss function config for traing/eval process
Loss:# loss function
Train:# list of training losses
-DistillationDMLLoss:# distillation DMLLoss. DMLLoss is encapsulated to support loss function of distillation (in dict)
weight:1.0# weight of loss
model_name_pairs:# model pair used to compute loss. Here loss function between Student and Teacher is computed
> Ying Zhang, Tao Xiang, Timothy M. Hospedales, Huchuan Lu
>
> CVPR, 2018
In the DML paper, the process of distillation does not depend on a teacher model. Two models with the same architecture learn from each other and calculate the KL divergence of each other's logits.
* Note: Complete PPLCNet_x2_5 The model have been trained for 360 epochs. For comparison, both baseline and DML have been trained for 100 epochs. Therefore, the accuracy is lower than the model (76.60%) opened on the official website.
##### 1.2.2.2 Configuration of DML
The DML configuration is shown below. In the `Arch` field, you need to define both the student model and the teacher model. Both models need to be updated. In the `Loss` field, you need to define `DistillationDMLLoss` (JS-Div between student and teacher) and `DistillationGTCELoss` (CE loss with ground truth labels) as the training loss.
```yaml
Arch:
name:"DistillationModel"
class_num:&class_num1000
pretrained_list:
freeze_params_list:# mutual learning, params of both models are not freezed
-False
-False
models:
-Teacher:
name:PPLCNet_x2_5# mutual learning, so pretrained models are not loaded for both models
class_num:*class_num
pretrained:False
-Student:
name:PPLCNet_x2_5
class_num:*class_num
pretrained:False
Loss:
Train:
-DistillationGTCELoss:# CE loss with ground truth labels needs to be computed for both models because pretrained models are not loaded
weight:1.0
model_names:["Student","Teacher"]
-DistillationDMLLoss:
weight:1.0
model_name_pairs:
-["Student","Teacher"]
Eval:
-CELoss:
weight:1.0
```
<aname='1.2.3'></a>
#### 1.2.3 UDML
##### 1.2.3.1 Introduction to UDML
UDML is a teacher-free knowledge distillation algorithm proposed by PaddleCV group. It is improved based on DML. In addition to the outputs, it also considers the middle layers features in the distillation process, so as to further improve the accuracy of knowledge distillation. For more information about UDML and its application, please refer to: [PP-ShiTu paper](https://arxiv.org/abs/2111.00775) and [PP-OCRv3 paper](https://arxiv.org/abs/2109.03144).
-DistillationDistanceLoss:# distance loss based on features. Here l2 loss is used to calculate the distance between block5s
weight:1.0
key:"blocks5"
model_name_pairs:
-["Student","Teacher"]
Eval:
-CELoss:
weight:1.0
```
**Note(:**`return_patterns` are specified in the network above. The function of returning middle layer features is based on TheseusLayer. For more information about usage of TheseusLayer, please refer to: [Usage of TheseusLayer](theseus_layer.md).
<aname='1.2.4'></a>
#### 1.2.4 AFD
##### 1.2.4.1 Introduction to AFD
Paper:
> [Show, attend and distill: Knowledge distillation via attention-based feature matching](https://arxiv.org/abs/2102.02973)
>
> Mingi Ji, Byeongho Heo, Sungrae Park
>
> AAAI, 2018
AFD proposes to use attention based meta network to learn the relative similarity between features in the distillation process, and apply the identified similarity relation to control the distillation intensity of all possible feature maps pairs.
Note: In order to keep alignment with the training configuration in the paper, the number of training iterations is set to be 100 epochs, so the baseline accuracy is lower than the open source model accuracy in PaddleClas (71.0%).
##### 1.2.4.2 Configuration of AFD
The AFD configuration is shown below. In the `Arch` field, you need to define both the student model and the teacher model. The teacher model has fixed parameters. In the `Loss` field, you need to define `DistillationKLDivLoss` (KL-Div between student and teacher), `AFDLoss` (AFD loss between student and teacher) and `DistillationGTCELoss` (CE loss with ground truth labels) as the training loss.
```yaml
Arch:
name:"DistillationModel"
pretrained_list:
freeze_params_list:
models:
-Teacher:
name:AttentionModel# contains several serial networks. The following networks take outputs of previous networks as input
pretrained_list:
freeze_params_list:
-True
-False
models:
# basic network of AttentionModel
-ResNet34:
name:ResNet34
pretrained:True
# return_patterns means that in addition to the output logits, the middle feature maps with the corresponding names will also be returned
-DistillationKLDivLoss:# distillation KL-Div loss, features are extracted according to names in model_name_pairs to calculate loss
weight:0.9# weight of loss
model_name_pairs:[["Student","Teacher"]]
temperature:4
key:logits
-AFDLoss:# AFD loss
weight:50.0
model_name_pair:["Student","Teacher"]
student_keys:["bilinear_key","value"]
teacher_keys:["query","value"]
s_shapes:*s_shapes
t_shapes:*t_shapes
Eval:
-CELoss:
weight:1.0
```
**Note(:**`return_patterns` are specified in the network above. The function of returning middle layer features is based on TheseusLayer. For more information about usage of TheseusLayer, please refer to: [Usage of TheseusLayer](theseus_layer.md).
DKD reformulates the classical KD loss into two parts, i.e., target class knowledge distillation (TCKD) and non-target class knowledge distillation (NCKD). The effect of the two parts is studied separately, and their weights can be adjusted independently, improving the accuracy and flexibility of distillation.
The DKD configuration is shown below. In the `Arch` field, you need to define both the student model and the teacher model. The teacher model has fixed parameters, and the pretrained parameters are loaded. In the `Loss` field, you need to define `DistillationDKDLoss` (DKD loss between student and teacher) and `DistillationGTCELoss` (CE loss with ground truth labels) as the training loss.
```yaml
Arch:
name:"DistillationModel"
# if not null, its lengths should be same as models
pretrained_list:
# if not null, its lengths should be same as models
freeze_params_list:
-True
-False
models:
-Teacher:
name:ResNet34
pretrained:True
-Student:
name:ResNet18
pretrained:False
infer_model_name:"Student"
# loss function config for traing/eval process
Loss:
Train:
-DistillationGTCELoss:
weight:1.0
model_names:["Student"]
-DistillationDKDLoss:
weight:1.0
model_name_pairs:[["Student","Teacher"]]
temperature:1
alpha:1.0
beta:1.0
Eval:
-CELoss:
weight:1.0
```
<aname='1.2.6'></a>
#### 1.2.6 DIST
##### 1.2.6.1 Introduction to DIST
Paper:
> [Knowledge Distillation from A Stronger Teacher](https://arxiv.org/pdf/2205.10536v1.pdf)
When using the KD method for distillation, as the accuracy of the teacher model is improved, the effect of distillation is often difficult to improve simultaneously. This paper proposes the DIST method, which uses the Pearson correlation coefficient to represent the difference between the student model and the teacher model, instead of the default KL-divergence in the distillation process, so as to ensure that the model can learn more accurate correlation information.
The DIST configuration is shown below. In the `Arch` field, you need to define both the student model and the teacher model. The teacher model has fixed parameters, and the pretrained parameters are loaded. In the `Loss` field, you need to define `DistillationDISTLoss` (DIST loss between student and teacher) and `DistillationGTCELoss` (CE loss with ground truth labels) as the training loss.
```yaml
Arch:
name:"DistillationModel"
# if not null, its lengths should be same as models
pretrained_list:
# if not null, its lengths should be same as models
This method performs distillation on the feature map. In the process of distillation, random masks are applied to the features, and the students are forced to use some features to generate all the features of the teacher model, so as to improve the representation ability of the student model. MGD achieve state-of-the-art performance on the feature distillation task, and has been widely verified to be effective in tasks such as detection and segmentation.
The MGD configuration is shown below. In the `Arch` field, you need to define both the student model and the teacher model. The teacher model has fixed parameters, and the pretrained parameters are loaded. In the `Loss` field, you need to define `DistillationPairLoss` (MGD loss between student and teacher) and `DistillationGTCELoss` (CE loss with ground truth labels) as the training loss.
```yaml
Arch:
name:"DistillationModel"
class_num:&class_num1000
# if not null, its lengths should be same as models
pretrained_list:
# if not null, its lengths should be same as models
Weighted Soft Labels (WSL) loss function assigns weights to the KD Loss of each sample according to the CE Loss ratio of the teacher model and the student model with respect to the ground-truth labels. If the student model predicts a certain sample better than the teacher model, a smaller weight will be assigned to the sample. The method is simple and effective. It enables the weight of each sample to be adaptively adjusted, thereby improving the distillation accuracy.
The WSL configuration is shown below. In the `Arch` field, you need to define both the student model and the teacher model. The teacher model has fixed parameters, and the pretrained parameters are loaded. In the `Loss` field, you need to define `DistillationWSLLoss` (WSL loss between student and teacher) and `DistillationGTCELoss` (CE loss with ground truth labels) as the training loss.
```yaml
# model architecture
Arch:
name:"DistillationModel"
# if not null, its lengths should be same as models
pretrained_list:
# if not null, its lengths should be same as models
freeze_params_list:
-True
-False
models:
-Teacher:
name:ResNet34
pretrained:True
-Student:
name:ResNet18
pretrained:False
infer_model_name:"Student"
# loss function config for traing/eval process
Loss:
Train:
-DistillationGTCELoss:
weight:1.0
model_names:["Student"]
-DistillationWSLLoss:
weight:2.5
model_name_pairs:[["Student","Teacher"]]
temperature:2
Eval:
-CELoss:
weight:1.0
```
<aname="2"></a>
## 2. Training, Evaluation and Prediction
<aname="2.1"></a>
### 2.1 Environment Configuration
* Installation: Please refer to [Paddle Installation Tutorial](../installation/install_paddle.md) and [PaddleClas Installation Tutorial](../../installation.md) to configure the running environment.
<aname="2.2"></a>
### 2.2 Data Preparation
Please prepare the ImageNet-1k dataset on [ImageNet website](https://www.image-net.org/).
Enter PaddleClas directory.
```
cd path_to_PaddleClas
```
Enter `dataset/` directory, name the downloaded data `ILSVRC2012` and store it here. The `ILSVRC2012` directory contains 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 training set and validation set, respectively. `train_list.txt` and `val_list.txt` are label files for training set and validation set, respectively.
If unlabeled data similar to the training set scenario is included, they can also be organized in the same way as the training set labels. Place the file in the same directory as the currently labeled dataset, and mark its label value as 0. Suppose the organized tag file is named `train_list_unlabel.txt`, you can use the following command to generate a label file for SSLD training.
* For more information about the format of `train_list.txt` and `val_list.txt`, you may refer to [Format Description of PaddleClas Classification Dataset](../single_label_classification/dataset.md#1-数据集格式说明) .
<aname="2.3"></a>
### 2.3 Model Training
In this section, the process of model training, evaluation and prediction of knowledge distillation algorithm will be introduced using the SSLD knowledge distillation algorithm as an example. The configuration file is [PPLCNet_x2_5_ssld.yaml](../../../../ppcls/configs/ImageNet/Distillation/PPLCNet_x2_5_ssld.yaml). You can use the following command to complete the model training.
where `-o Global.pretrained_model="output/DistillationModel/best_model"` specifies the path of the current optimal weights. If you need to specify other weights, you can simply replace the path.
<aname="2.5"></a>
### 2.5 Model Prediction
After training is completed, the trained model can be loaded for prediction. A complete example is provided in `tools/infer.py`. You can use the model for prediction by executing the following command:
* Here `-o Global.pretrained_model="output/ResNet50/best_model"` specifies the path of the current optimal weights. If you need to specify other weights, you can simply replace the path.
* Image `docs/images/inference_deployment/whl_demo.jpg` is predicted by default. You can also predict other images by adding a field `-o Infer.infer_imgs=xxx`.
<aname="2.6"></a>
### 2.6 Model Export & Inference
PaddleInference is a native inference library for PaddlePaddle, which can be used on servers and clouds to provide high-performance inference. PaddleInference can use MKLDNN, CUDNN, and TensorRT to accelerate model inference, thereby achieving better performance compared with inference based directly on the trained model. For more information about PaddleInference, please refer to [Paddle Inference Tutorial](https://www.paddlepaddle.org.cn/documentation/docs/zh/guides/infer/inference/inference_cn.html).
The model needs to be exported before inference. For models obtained from knowledge distillation, `-o Global.infer_model_name=Student` should be specified when exporting to indicate the model to be exported is the student model. The complete command is shown below.
3 files will be generated in `inference` directory: `inference.pdiparams`, `inference.pdiparams.info` and `inference.pdmodel`.
For more information about model inference, please refer to: [Python Inference](../../deployment/image_classification/python.md).
<aname="3"></a>
## 3. References
[1] Hinton G, Vinyals O, Dean J. Distilling the knowledge in a neural network[J]. arXiv preprint arXiv:1503.02531, 2015.
[2] Bagherinezhad H, Horton M, Rastegari M, et al. Label refinery: Improving imagenet classification through label progression[J]. arXiv preprint arXiv:1805.02641, 2018.
[3] Yalniz I Z, Jégou H, Chen K, et al. Billion-scale semi-supervised learning for image classification[J]. arXiv preprint arXiv:1905.00546, 2019.
[4] Cubuk E D, Zoph B, Mane D, et al. Autoaugment: Learning augmentation strategies from data[C]//Proceedings of the IEEE conference on computer vision and pattern recognition. 2019: 113-123.
[5] Touvron H, Vedaldi A, Douze M, et al. Fixing the train-test resolution discrepancy[C]//Advances in Neural Information Processing Systems. 2019: 8250-8260.
[6] Cui C, Guo R, Du Y, et al. Beyond Self-Supervision: A Simple Yet Effective Network Distillation Alternative to Improve Backbones[J]. arXiv preprint arXiv:2103.05959, 2021.
[7] Zhang Y, Xiang T, Hospedales T M, et al. Deep mutual learning[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2018: 4320-4328.
[8] Heo B, Kim J, Yun S, et al. A comprehensive overhaul of feature distillation[C]//Proceedings of the IEEE/CVF International Conference on Computer Vision. 2019: 1921-1930.
[9] Du Y, Li C, Guo R, et al. PP-OCRv2: Bag of Tricks for Ultra Lightweight OCR System[J]. arXiv preprint arXiv:2109.03144, 2021.
[10] Park W, Kim D, Lu Y, et al. Relational knowledge distillation[C]//Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. 2019: 3967-3976.
[11] Zhao B, Cui Q, Song R, et al. Decoupled Knowledge Distillation[J]. arXiv preprint arXiv:2203.08679, 2022.
[12] Ji M, Heo B, Park S. Show, attend and distill: Knowledge distillation via attention-based feature matching[C]//Proceedings of the AAAI Conference on Artificial Intelligence. 2021, 35(9): 7945-7952.
[13] Huang T, You S, Wang F, et al. Knowledge Distillation from A Stronger Teacher[J]. arXiv preprint arXiv:2205.10536, 2022.
At present, although the lightweight models are so plentiful, there are few models specially optimized for Intel CPU platform. We have proposed [PPLCNetV1](PP-LCNet.md) , which pursues faster inference speed, so its performance is limited and the performance is insufficient when dealing with complex tasks. Therefore, we now propose PPLCNetV2 to fill the gap in the lack of a higher performance model on CPU platform. For the consideration of compatibility, OpenVINO is currently a widely used deployment framework in Intel CPU platform. Therefore, we focus on optimizing the model for the deployment scheme of Intel CPU with OpenVINO.
<aname="2"></a>
## 2. Methods
The overview of PPLCNetV2 is shown in the figure above. PPLCNetV2 is obtained on PPLCNet. The details of optimization tracks is shown in this section.
<aname="2.1"></a>
### 2.1 Re-parameterization
There are lots of depthwise convolution in PPLCNetV2, so we optimize some of the depthwise convolution by the re-parameterization. The size of the convolution kernel affects the size of the model's receptive field, which affects the model's ability to capture more global or local features. In order to help the model build different scales features, we use 5\*5, 3\*3 and 1\*1 size convolution kernel. The details is shown in the figure below.
<aname="2.2"></a>
### 2.2 PW Conv
We know that the network is more deeper, the model is more stronger. So we replaced some point convolution layer with two layers that squeeze and expand the channel dimensions of the feature, respectively. The details is shown in the figure below. Finally, we use this optimization method in the second last-to-last block.
<aname="2.3"></a>
### 2.3 Shortcut
It is believed that the Shortcut can alleviate the vanishing gradient problem, so it is more important for the improvement of deep networks. However, Shortcut is generally used with caution in the lightweight models because it results in an elementwise addition operation and possibly memory access. We experimented on the influence of shortcut on the model at different stage. Finally, we only used Shortcut in the last block, as shown in the figure below.
<aname="2.4"></a>
### 2.4 Activation Function
In recent years, many activation functions have been proposed, such as ReLU, Swish, Hard-Swish, Mish, but they usually pay more attention to the improvement of model performance without considering the impact on model efficiency. In particular, some activation functions contain complex operations, or are difficult to the optimization of the inference platform. So some act functions can have a serious efficiency impact in model. We experimented with different activation functions to evaluate their efficiency. We found that although the ReLU activation function is slightly lower than H-Swish in model performance, the gap is very small in larger models of this magnitude, but ReLU has a huge advantage in speed because which is so brief that very easy to optimize. There is no doubt that we chose the ReLU activation function after considering it all.
<aname="2.5"></a>
### 2.5 SE
SE Module has been concerned since it was proposed and has almost become a necessary option for model designs. It can significantly improve model performance by helping the model to improve the channel attention ability. However, in the lightweight models, SE Module will not only improve the performance of the model, but also increase the inference latency. In PPLCNetV2, we used SE more sparingly, only on the penultimate blocks.
<aname="3"></a>
## 3. Experiments
The accuracy on ImageNet1k, latency of inference and download links of pretrained models are provided as follows.
| Model | Params(M) | FLOPs(M) | Top-1 Acc(\%) | Top-5 Acc(\%) | Latency(ms) | download links of pretrained model | download links of inference model |
* 1. where `_ssld` represents the model after using `SSLD distillation`. For details about `SSLD distillation`, see [SSLD distillation](../advanced_tutorials/distillation/distillation_en.md);
* 2. The latency is tested with officially OpenVINO benchmark tool on Intel(R) Xeon(R) Gold 6271C CPU @ 2.60GHz.
