---
comments: true
---

# PaddleX模型列表（海光 DCU）

PaddleX 内置了多条产线，每条产线都包含了若干模块，每个模块包含若干模型，具体使用哪些模型，您可以根据下边的 benchmark 数据来选择。如您更考虑模型精度，请选择精度较高的模型，如您更考虑模型存储大小，请选择存储大小较小的模型。

## 图像分类模块
<table>
<thead>
<tr>
<th>模型名称</th>
<th>Top1 Acc（%）</th>
<th>模型存储大小（M)</th>
<th>模型下载链接</th></tr>
</thead>
<tbody>
<tr>
<td>ResNet18</td>
<td>71.0</td>
<td>41.5 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/ResNet18_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ResNet18_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>ResNet34</td>
<td>74.6</td>
<td>77.3 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/ResNet34_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ResNet34_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>ResNet50</td>
<td>76.5</td>
<td>90.8 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/ResNet50_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ResNet50_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>ResNet101</td>
<td>77.6</td>
<td>158.7 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/ResNet101_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ResNet101_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>ResNet152</td>
<td>78.3</td>
<td>214.2 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/ResNet152_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/ResNet152_pretrained.pdparams">训练模型</a></td></tr>
</tbody>
</table>
<b>注：以上精度指标为</b>[ImageNet-1k](https://www.image-net.org/index.php)<b>验证集 Top1 Acc。</b>

## [图像多标签分类模块](../module_usage/tutorials/cv_modules/image_multilabel_classification.md)
<table>
<thead>
<tr>
<th>模型名称</th>
<th>mAP（%）</th>
<th>模型存储大小</th>
<th>模型下载链接</th></tr>
</thead>
<tbody>
<tr>
<td>CLIP_vit_base_patch16_448_ML</td>
<td>89.15</td>
<td>325.6 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/CLIP_vit_base_patch16_448_ML_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/CLIP_vit_base_patch16_448_ML_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>PP-HGNetV2-B0_ML</td>
<td>80.98</td>
<td>39.6 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-HGNetV2-B0_ML_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-HGNetV2-B0_ML_pretrained.pdparams">训练模型</a></td>
<tr>
<td>PP-HGNetV2-B4_ML</td>
<td>87.96</td>
<td>88.5 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-HGNetV2-B4_ML_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-HGNetV2-B4_ML_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>PP-HGNetV2-B6_ML</td>
<td>91.06</td>
<td>286.5 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-HGNetV2-B6_ML_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-HGNetV2-B6_ML_pretrained.pdparams">训练模型</a></td></tr>
</tbody>
</table>
<b>注：以上精度指标为 [COCO2017](https://cocodataset.org/#home) 的多标签分类任务mAP。</b>

## [图像特征模块](../module_usage/tutorials/cv_modules/image_feature.md)
<table>
<thead>
<tr>
<th>模型名称</th>
<th>recall@1（%）</th>
<th>模型存储大小</th>
<th>模型下载链接</th></tr>
</thead>
<tbody>
<tr>
<td>PP-ShiTuV2_rec_CLIP_vit_base</td>
<td>88.69</td>
<td>306.6 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-ShiTuV2_rec_CLIP_vit_base_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-ShiTuV2_rec_CLIP_vit_base_pretrained.pdparams">训练模型</a></td></tr>
</tbody>
</table>
<b>注：以上精度指标为 AliProducts recall@1。</b>

## 目标检测模块
<table>
<thead>
<tr>
<th>模型名称</th>
<th>mAP（%）</th>
<th>模型存储大小（M)</th>
<th>模型下载链接</th></tr>
</thead>
<tbody>
<tr>
<td>PicoDet-L</td>
<td>42.6</td>
<td>20.9 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PicoDet-L_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PicoDet-L_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>PicoDet-M</td>
<td>37.5</td>
<td>16.8 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PicoDet-M_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PicoDet-M_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>PicoDet-S</td>
<td>29.1</td>
<td>4.4 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PicoDet-S_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PicoDet-S_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>PicoDet-XS</td>
<td>26.2</td>
<td>5.7M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PicoDet-XS_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PicoDet-XS_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>PP-YOLOE_plus-L</td>
<td>52.9</td>
<td>185.3 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-YOLOE_plus-L_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-YOLOE_plus-L_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>PP-YOLOE_plus-M</td>
<td>49.8</td>
<td>83.2 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-YOLOE_plus-M_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-YOLOE_plus-M_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>PP-YOLOE_plus-S</td>
<td>43.7</td>
<td>28.3 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-YOLOE_plus-S_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-YOLOE_plus-S_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>PP-YOLOE_plus-X</td>
<td>54.7</td>
<td>349.4 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-YOLOE_plus-X_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-YOLOE_plus-X_pretrained.pdparams">训练模型</a></td></tr>
</tbody>
</table>
<b>注：以上精度指标为</b>[COCO2017](https://cocodataset.org/#home)<b>验证集 mAP(0.5:0.95)。</b>

## [小目标检测模块](../module_usage/tutorials/cv_modules/small_object_detection.md)
<table>
<thead>
<tr>
<th>模型名称</th>
<th>mAP（%）</th>
<th>模型存储大小</th>
<th>模型下载链接</th></tr>
</thead>
<tbody>
<tr>
<td>PP-YOLOE_plus_SOD-S</td>
<td>25.1</td>
<td>77.3 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-YOLOE_plus_SOD-S_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-YOLOE_plus_SOD-S_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>PP-YOLOE_plus_SOD-L</td>
<td>31.9</td>
<td>325.0 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-YOLOE_plus_SOD-L_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-YOLOE_plus_SOD-L_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>PP-YOLOE_plus_SOD-largesize-L</td>
<td>42.7</td>
<td>340.5 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-YOLOE_plus_SOD-largesize-L_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-YOLOE_plus_SOD-largesize-L_pretrained.pdparams">训练模型</a></td></tr>
</tbody>
</table>
<b>注：以上精度指标为 </b>[VisDrone-DET](https://github.com/VisDrone/VisDrone-Dataset)<b> 验证集 mAP(0.5:0.95)。</b>

## 语义分割模块
<table>
<thead>
<tr>
<th>模型名称</th>
<th>mloU（%）</th>
<th>模型存储大小（M)</th>
<th>模型下载链接</th></tr>
</thead>
<tbody>
<tr>
<td>Deeplabv3_Plus-R50</td>
<td>80.36</td>
<td>94.9 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/Deeplabv3_Plus-R50_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/Deeplabv3_Plus-R50_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>Deeplabv3_Plus-R101</td>
<td>81.10</td>
<td>162.5 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/Deeplabv3_Plus-R101_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/Deeplabv3_Plus-R101_pretrained.pdparams">训练模型</a></td></tr>
</tbody>
</table>
<b>注：以上精度指标为</b>[Cityscapes](https://www.cityscapes-dataset.com/)<b>数据集 mloU。</b>

## [异常检测模块](../module_usage/tutorials/cv_modules/anomaly_detection.md)
<table>
<thead>
<tr>
<th>模型名称</th>
<th>Avg（%）</th>
<th>模型存储大小</th>
<th>模型下载链接</th></tr>
</thead>
<tbody>
<tr>
<td>STFPM</td>
<td>96.2</td>
<td>21.5 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/STFPM_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/STFPM_pretrained.pdparams">训练模型</a></td></tr>
</tbody>
</table>
<b>注：以上精度指标为 </b>[MVTec AD](https://www.mvtec.com/company/research/datasets/mvtec-ad)<b> 验证集 平均异常分数。</b>

## [人脸检测模块](../module_usage/tutorials/cv_modules/face_detection.md)
<table>
<thead>
<tr>
<th>模型名称</th>
<th style="text-align: center;">AP (%)<br/>Easy/Medium/Hard</th>
<th>模型存储大小</th>
<th>模型下载链接</th></tr>
</thead>
<tbody>
<tr>
<td>PicoDet_LCNet_x2_5_face</td>
<td style="text-align: center;">93.7/90.7/68.1</td>
<td>28.9 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PicoDet_LCNet_x2_5_face_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PicoDet_LCNet_x2_5_face_pretrained.pdparams">训练模型</a></td></tr>
</tbody>
</table>
**注：以上精度指标是在WIDER-FACE验证集上，以640
\*640作为输入尺寸评估得到的。**

## 文本检测模块
<table>
<thead>
<tr>
<th>模型名称</th>
<th>检测Hmean（%）</th>
<th>模型存储大小（M)</th>
<th>模型下载链接</th></tr>
</thead>
<tbody>
<tr>
<td>PP-OCRv4_mobile_det</td>
<td>77.79</td>
<td>4.2 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-OCRv4_mobile_det_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-OCRv4_mobile_det_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>PP-OCRv4_server_det</td>
<td>82.69</td>
<td>100.1M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-OCRv4_server_det_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-OCRv4_server_det_pretrained.pdparams">训练模型</a></td></tr>
</tbody>
</table>
<b>注：以上精度指标的评估集是 PaddleOCR 自建的中文数据集，覆盖街景、网图、文档、手写多个场景，其中检测包含 500 张图片。</b>

## 文本识别模块
<table>
<thead>
<tr>
<th>模型名称</th>
<th>识别Avg Accuracy(%)</th>
<th>模型存储大小（M)</th>
<th>模型下载链接</th></tr>
</thead>
<tbody>
<tr>
<td>PP-OCRv4_mobile_rec</td>
<td>78.20</td>
<td>10.6 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-OCRv4_mobile_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-OCRv4_mobile_rec_pretrained.pdparams">训练模型</a></td></tr>
<tr>
<td>PP-OCRv4_server_rec</td>
<td>79.20</td>
<td>71.2 M</td>
<td><a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_inference_model/paddle3.0rc0/PP-OCRv4_server_rec_infer.tar">推理模型</a>/<a href="https://paddle-model-ecology.bj.bcebos.com/paddlex/official_pretrained_model/PP-OCRv4_server_rec_pretrained.pdparams">训练模型</a></td></tr>
</tbody>
</table>
<b>注：以上精度指标的评估集是 PaddleOCR 自建的中文数据集，覆盖街景、网图、文档、手写多个场景，其中文本识别包含 1.1w 张图片。</b>