models_list.md 38 KB
简体中文 | English
PaddleX模型列表（CPU/GPU）
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，如您更考虑模型推理速度，请选择推理速度较快的模型，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
|模型名称|Top1 Acc（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |CLIP_vit_base_patch16_224|85.36|13.1957|285.493|306.5 M|CLIP_vit_base_patch16_224.yaml| |CLIP_vit_large_patch14_224|88.1|51.1284|1131.28|1.04 G|CLIP_vit_large_patch14_224.yaml| |ConvNeXt_base_224|83.84|12.8473|1513.87|313.9 M|ConvNeXt_base_224.yaml| |ConvNeXt_base_384|84.90|31.7607|3967.05|313.9 M|ConvNeXt_base_384.yaml| |ConvNeXt_large_224|84.26|26.8103|2463.56|700.7 M|ConvNeXt_large_224.yaml| |ConvNeXt_large_384|85.27|66.4058|6598.92|700.7 M|ConvNeXt_large_384.yaml| |ConvNeXt_small|83.13|9.74075|1127.6|178.0 M|ConvNeXt_small.yaml| |ConvNeXt_tiny|82.03|5.48923|672.559|101.4 M|ConvNeXt_tiny.yaml| |FasterNet-L|83.5|23.4415|-|357.1 M|FasterNet-L.yaml| |FasterNet-M|83.0|21.8936|-|204.6 M|FasterNet-M.yaml| |FasterNet-S|81.3|13.0409|-|119.3 M|FasterNet-S.yaml| |FasterNet-T0|71.9|12.2432|-|15.1 M|FasterNet-T0.yaml| |FasterNet-T1|75.9|11.3562|-|29.2 M|FasterNet-T1.yaml| |FasterNet-T2|79.1|10.703|-|57.4 M|FasterNet-T2.yaml| |MobileNetV1_x0_5|63.5|1.86754|7.48297|4.8 M|MobileNetV1_x0_5.yaml| |MobileNetV1_x0_25|51.4|1.83478|4.83674|1.8 M|MobileNetV1_x0_25.yaml| |MobileNetV1_x0_75|68.8|2.57903|10.6343|9.3 M|MobileNetV1_x0_75.yaml| |MobileNetV1_x1_0|71.0|2.78781|13.98|15.2 M|MobileNetV1_x1_0.yaml| |MobileNetV2_x0_5|65.0|4.94234|11.1629|7.1 M|MobileNetV2_x0_5.yaml| |MobileNetV2_x0_25|53.2|4.50856|9.40991|5.5 M|MobileNetV2_x0_25.yaml| |MobileNetV2_x1_0|72.2|6.12159|16.0442|12.6 M|MobileNetV2_x1_0.yaml| |MobileNetV2_x1_5|74.1|6.28385|22.5129|25.0 M|MobileNetV2_x1_5.yaml| |MobileNetV2_x2_0|75.2|6.12888|30.8612|41.2 M|MobileNetV2_x2_0.yaml| |MobileNetV3_large_x0_5|69.2|6.31302|14.5588|9.6 M|MobileNetV3_large_x0_5.yaml| |MobileNetV3_large_x0_35|64.3|5.76207|13.9041|7.5 M|MobileNetV3_large_x0_35.yaml| |MobileNetV3_large_x0_75|73.1|8.41737|16.9506|14.0 M|MobileNetV3_large_x0_75.yaml| |MobileNetV3_large_x1_0|75.3|8.64112|19.1614|19.5 M|MobileNetV3_large_x1_0.yaml| |MobileNetV3_large_x1_25|76.4|8.73358|22.1296|26.5 M|MobileNetV3_large_x1_25.yaml| |MobileNetV3_small_x0_5|59.2|5.16721|11.2688|6.8 M|MobileNetV3_small_x0_5.yaml| |MobileNetV3_small_x0_35|53.0|5.22053|11.0055|6.0 M|MobileNetV3_small_x0_35.yaml| |MobileNetV3_small_x0_75|66.0|5.39831|12.8313|8.5 M|MobileNetV3_small_x0_75.yaml| |MobileNetV3_small_x1_0|68.2|6.00993|12.9598|10.5 M|MobileNetV3_small_x1_0.yaml| |MobileNetV3_small_x1_25|70.7|6.9589|14.3995|13.0 M|MobileNetV3_small_x1_25.yaml| |MobileNetV4_conv_large|83.4|12.5485|51.6453|125.2 M|MobileNetV4_conv_large.yaml| |MobileNetV4_conv_medium|79.9|9.65509|26.6157|37.6 M|MobileNetV4_conv_medium.yaml| |MobileNetV4_conv_small|74.6|5.24172|11.0893|14.7 M|MobileNetV4_conv_small.yaml| |MobileNetV4_hybrid_large|83.8|20.0726|213.769|145.1 M|MobileNetV4_hybrid_large.yaml| |MobileNetV4_hybrid_medium|80.5|19.7543|62.2624|42.9 M|MobileNetV4_hybrid_medium.yaml| |PP-HGNet_base|85.0|14.2969|327.114|249.4 M|PP-HGNet_base.yaml| |PP-HGNet_small|81.51|5.50661|119.041|86.5 M|PP-HGNet_small.yaml| |PP-HGNet_tiny|79.83|5.22006|69.396|52.4 M|PP-HGNet_tiny.yaml| |PP-HGNetV2-B0|77.77|6.53694|23.352|21.4 M|PP-HGNetV2-B0.yaml| |PP-HGNetV2-B1|79.18|6.56034|27.3099|22.6 M|PP-HGNetV2-B1.yaml| |PP-HGNetV2-B2|81.74|9.60494|43.1219|39.9 M|PP-HGNetV2-B2.yaml| |PP-HGNetV2-B3|82.98|11.0042|55.1367|57.9 M|PP-HGNetV2-B3.yaml| |PP-HGNetV2-B4|83.57|9.66407|54.2462|70.4 M|PP-HGNetV2-B4.yaml| |PP-HGNetV2-B5|84.75|15.7091|115.926|140.8 M|PP-HGNetV2-B5.yaml| |PP-HGNetV2-B6|86.30|21.226|255.279|268.4 M|PP-HGNetV2-B6.yaml| |PP-LCNet_x0_5|63.14|3.67722|6.66857|6.7 M|PP-LCNet_x0_5.yaml| |PP-LCNet_x0_25|51.86|2.65341|5.81357|5.5 M|PP-LCNet_x0_25.yaml| |PP-LCNet_x0_35|58.09|2.7212|6.28944|5.9 M|PP-LCNet_x0_35.yaml| |PP-LCNet_x0_75|68.18|3.91032|8.06953|8.4 M|PP-LCNet_x0_75.yaml| |PP-LCNet_x1_0|71.32|3.84845|9.23735|10.5 M|PP-LCNet_x1_0.yaml| |PP-LCNet_x1_5|73.71|3.97666|12.3457|16.0 M|PP-LCNet_x1_5.yaml| |PP-LCNet_x2_0|75.18|4.07556|16.2752|23.2 M|PP-LCNet_x2_0.yaml| |PP-LCNet_x2_5|76.60|4.06028|21.5063|32.1 M|PP-LCNet_x2_5.yaml| |PP-LCNetV2_base|77.05|5.23428|19.6005|23.7 M|PP-LCNetV2_base.yaml| |PP-LCNetV2_large |78.51|6.78335|30.4378|37.3 M|PP-LCNetV2_large.yaml| |PP-LCNetV2_small|73.97|3.89762|13.0273|14.6 M|PP-LCNetV2_small.yaml| |ResNet18_vd|72.3|3.53048|31.3014|41.5 M|ResNet18_vd.yaml| |ResNet18|71.0|2.4868|27.4601|41.5 M|ResNet18.yaml| |ResNet34_vd|76.0|5.60675|56.0653|77.3 M|ResNet34_vd.yaml| |ResNet34|74.6|4.16902|51.925|77.3 M|ResNet34.yaml| |ResNet50_vd|79.1|10.1885|68.446|90.8 M|ResNet50_vd.yaml| |ResNet50|76.5|9.62383|64.8135|90.8 M|ResNet50.yaml| |ResNet101_vd|80.2|20.0563|124.85|158.4 M|ResNet101_vd.yaml| |ResNet101|77.6|19.2297|121.006|158.7 M|ResNet101.yaml| |ResNet152_vd|80.6|29.6439|181.678|214.3 M|ResNet152_vd.yaml| |ResNet152|78.3|30.0461|177.707|214.2 M|ResNet152.yaml| |ResNet200_vd|80.9|39.1628|235.185|266.0 M|ResNet200_vd.yaml| |StarNet-S1|73.6|9.895|23.0465|11.2 M|StarNet-S1.yaml| |StarNet-S2|74.8|7.91279|21.9571|14.3 M|StarNet-S2.yaml| |StarNet-S3|77.0|10.7531|30.7656|22.2 M|StarNet-S3.yaml| |StarNet-S4|79.0|15.2868|43.2497|28.9 M|StarNet-S4.yaml| |SwinTransformer_base_patch4_window7_224|83.37|16.9848|383.83|310.5 M|SwinTransformer_base_patch4_window7_224.yaml| |SwinTransformer_base_patch4_window12_384|84.17|37.2855|1178.63|311.4 M|SwinTransformer_base_patch4_window12_384.yaml| |SwinTransformer_large_patch4_window7_224|86.19|27.5498|689.729|694.8 M|SwinTransformer_large_patch4_window7_224.yaml| |SwinTransformer_large_patch4_window12_384|87.06|74.1768|2105.22|696.1 M|SwinTransformer_large_patch4_window12_384.yaml| |SwinTransformer_small_patch4_window7_224|83.21|16.3982|285.56|175.6 M|SwinTransformer_small_patch4_window7_224.yaml| |SwinTransformer_tiny_patch4_window7_224|81.10|8.54846|156.306|100.1 M|SwinTransformer_tiny_patch4_window7_224.yaml|
**注：以上精度指标为 ImageNet-1k 验证集 Top1 Acc。**
图像多标签分类模块
|模型名称|mAP（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |CLIP_vit_base_patch16_448_ML|89.15|-|-|325.6 M|CLIP_vit_base_patch16_448_ML.yaml| |PP-HGNetV2-B0_ML|80.98|-|-|39.6 M|PP-HGNetV2-B0_ML.yaml| |PP-HGNetV2-B4_ML|87.96|-|-|88.5 M|PP-HGNetV2-B4_ML.yaml| |PP-HGNetV2-B6_ML|91.25|-|-|286.5 M|PP-HGNetV2-B6_ML.yaml| |PP-LCNet_x1_0_ML|77.96|-|-|29.4 M|PP-LCNet_x1_0_ML.yaml| |ResNet50_ML|83.50|-|-|108.9 M|ResNet50_ML.yaml|
注：以上精度指标为 COCO2017 的多标签分类任务mAP。
行人属性模块
|模型名称|mA（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |PP-LCNet_x1_0_pedestrian_attribute|92.2|3.84845|9.23735|6.7 M |PP-LCNet_x1_0_pedestrian_attribute.yaml|
注：以上精度指标为 PaddleX 内部自建数据集mA。
车辆属性模块
|模型名称|mA（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |PP-LCNet_x1_0_vehicle_attribute|91.7|3.84845|9.23735|6.7 M|PP-LCNet_x1_0_vehicle_attribute.yaml|
注：以上精度指标为 VeRi 数据集 mA。
图像特征模块
|模型名称|recall@1（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |PP-ShiTuV2_rec|84.2|5.23428|19.6005|16.3 M|PP-ShiTuV2_rec.yaml| |PP-ShiTuV2_rec_CLIP_vit_base|88.69|13.1957|285.493|306.6 M|PP-ShiTuV2_rec_CLIP_vit_base.yaml| |PP-ShiTuV2_rec_CLIP_vit_large|91.03|51.1284|1131.28|1.05 G|PP-ShiTuV2_rec_CLIP_vit_large.yaml|
注：以上精度指标为 AliProducts recall@1。
文档方向分类模块
|模型名称|Top-1 Acc（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |PP-LCNet_x1_0_doc_ori|99.26|3.84845|9.23735|7.1 M|PP-LCNet_x1_0_doc_ori.yaml|
注：以上精度指标为 PaddleX 内部自建数据集 Top-1 Acc 。
人脸特征模块
| 模型名称 | 输出特征维度 | Acc (%) AgeDB-30/CFP-FP/LFW |
GPU推理耗时 (ms) | CPU推理耗时 | 模型存储大小 (M) | yaml 文件 |
|---|---|---|---|---|---|---|
| MobileFaceNet | 128 | 96.28/96.71/99.58 | 4.1 | MobileFaceNet.yaml | ||
| ResNet50_face | 512 | 98.12/98.56/99.77 | 87.2 | ResNet50_face.yaml |
注：以上精度指标是分别在AgeDB-30、CFP-FP和LFW数据集上测得的Accuracy。
主体检测模块
|模型名称|mAP（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |PP-ShiTuV2_det|41.5|33.7426|537.003|27.6 M|PP-ShiTuV2_det.yaml|
注：以上精度指标为 PaddleClas主体检测数据集 mAP(0.5:0.95)。
目标检测模块
|模型名称|mAP（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |Cascade-FasterRCNN-ResNet50-FPN|41.1|-|-|245.4 M|Cascade-FasterRCNN-ResNet50-FPN.yaml| |Cascade-FasterRCNN-ResNet50-vd-SSLDv2-FPN|45.0|-|-|246.2 M|Cascade-FasterRCNN-ResNet50-vd-SSLDv2-FPN.yaml| |CenterNet-DLA-34|37.6|-|-|75.4 M|CenterNet-DLA-34.yaml| |CenterNet-ResNet50|38.9|-|-|319.7 M|CenterNet-ResNet50.yaml| |DETR-R50|42.3|59.2132|5334.52|159.3 M|DETR-R50.yaml| |FasterRCNN-ResNet34-FPN|37.8|-|-|137.5 M|FasterRCNN-ResNet34-FPN.yaml| |FasterRCNN-ResNet50-FPN|38.4|-|-|148.1 M|FasterRCNN-ResNet50-FPN.yaml| |FasterRCNN-ResNet50-vd-FPN|39.5|-|-|148.1 M|FasterRCNN-ResNet50-vd-FPN.yaml| |FasterRCNN-ResNet50-vd-SSLDv2-FPN|41.4|-|-|148.1 M|FasterRCNN-ResNet50-vd-SSLDv2-FPN.yaml| |FasterRCNN-ResNet50|36.7|-|-|120.2 M|FasterRCNN-ResNet50.yaml| |FasterRCNN-ResNet101-FPN|41.4|-|-|216.3 M|FasterRCNN-ResNet101-FPN.yaml| |FasterRCNN-ResNet101|39.0|-|-|188.1 M|FasterRCNN-ResNet101.yaml| |FasterRCNN-ResNeXt101-vd-FPN|43.4|-|-|360.6 M|FasterRCNN-ResNeXt101-vd-FPN.yaml| |FasterRCNN-Swin-Tiny-FPN|42.6|-|-|159.8 M|FasterRCNN-Swin-Tiny-FPN.yaml| |FCOS-ResNet50|39.6|103.367|3424.91|124.2 M|FCOS-ResNet50.yaml| |PicoDet-L|42.6|16.6715|169.904|20.9 M|PicoDet-L.yaml| |PicoDet-M|37.5|16.2311|71.7257|16.8 M|PicoDet-M.yaml| |PicoDet-S|29.1|14.097|37.6563|4.4 M |PicoDet-S.yaml| |PicoDet-XS|26.2|13.8102|48.3139|5.7M |PicoDet-XS.yaml| |PP-YOLOE_plus-L|52.9|33.5644|814.825|185.3 M|PP-YOLOE_plus-L.yaml| |PP-YOLOE_plus-M|49.8|19.843|449.261|83.2 M|PP-YOLOE_plus-M.yaml| |PP-YOLOE_plus-S|43.7|16.8884|223.059|28.3 M|PP-YOLOE_plus-S.yaml| |PP-YOLOE_plus-X|54.7|57.8995|1439.93|349.4 M|PP-YOLOE_plus-X.yaml| |RT-DETR-H|56.3|114.814|3933.39|435.8 M|RT-DETR-H.yaml| |RT-DETR-L|53.0|34.5252|1454.27|113.7 M|RT-DETR-L.yaml| |RT-DETR-R18|46.5|19.89|784.824|70.7 M|RT-DETR-R18.yaml| |RT-DETR-R50|53.1|41.9327|1625.95|149.1 M|RT-DETR-R50.yaml| |RT-DETR-X|54.8|61.8042|2246.64|232.9 M|RT-DETR-X.yaml| |YOLOv3-DarkNet53|39.1|40.1055|883.041|219.7 M|YOLOv3-DarkNet53.yaml| |YOLOv3-MobileNetV3|31.4|18.6692|267.214|83.8 M|YOLOv3-MobileNetV3.yaml| |YOLOv3-ResNet50_vd_DCN|40.6|31.6276|856.047|163.0 M|YOLOv3-ResNet50_vd_DCN.yaml| |YOLOX-L|50.1|185.691|1250.58|192.5 M|YOLOX-L.yaml| |YOLOX-M|46.9|123.324|688.071|90.0 M|YOLOX-M.yaml| |YOLOX-N|26.1|79.1665|155.59|3.4M|YOLOX-N.yaml| |YOLOX-S|40.4|184.828|474.446|32.0 M|YOLOX-S.yaml| |YOLOX-T|32.9|102.748|212.52|18.1 M|YOLOX-T.yaml| |YOLOX-X|51.8|227.361|2067.84|351.5 M|YOLOX-X.yaml|
**注：以上精度指标为 COCO2017 验证集 mAP(0.5:0.95)。**
小目标检测模块
|模型名称|mAP（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |PP-YOLOE_plus_SOD-S|25.1|65.4608|324.37|77.3 M|PP-YOLOE_plus_SOD-S.yaml| |PP-YOLOE_plus_SOD-L|31.9|57.1448|1006.98|325.0 M|PP-YOLOE_plus_SOD-L.yaml| |PP-YOLOE_plus_SOD-largesize-L|42.7|458.521|11172.7|340.5 M|PP-YOLOE_plus_SOD-largesize-L.yaml|
**注：以上精度指标为 VisDrone-DET 验证集 mAP(0.5:0.95)。**
行人检测模块
|模型名称|mAP（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |PP-YOLOE-L_human|48.0|32.7754|777.691|196.1 M|PP-YOLOE-L_human.yaml| |PP-YOLOE-S_human|42.5|15.0118|179.317|28.8 M|PP-YOLOE-S_human.yaml|
**注：以上精度指标为 CrowdHuman 验证集 mAP(0.5:0.95)。**
车辆检测模块
|模型名称|mAP（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |PP-YOLOE-L_vehicle|63.9|32.5619|775.633|196.1 M|PP-YOLOE-L_vehicle.yaml| |PP-YOLOE-S_vehicle|61.3|15.3787|178.441|28.8 M|PP-YOLOE-S_vehicle.yaml|
**注：以上精度指标为 PPVehicle 验证集 mAP(0.5:0.95)。**
人脸检测模块
|模型名称|AP (%)
Easy/Medium/Hard|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件|
|-|:-:|-|-|-|-|
| BlazeFace | 77.7/73.4/49.5 | | | 0.447 M | BlazeFace.yaml|
| BlazeFace-FPN-SSH | 83.2/80.5/60.5 | | | 0.606 M | BlazeFace-FPN-SSH.yaml |
| PicoDet_LCNet_x2_5_face | 93.7/90.7/68.1 | | | 28.9 M | PicoDet_LCNet_x2_5_face.yaml |
| PP-YOLOE_plus-S_face | 93.9/91.8/79.8 | | | 26.5 M |PP-YOLOE_plus-S_face |
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。
异常检测模块
|模型名称|Avg（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |STFPM|96.2|-|-|21.5 M|STFPM.yaml|
**注：以上精度指标为 MVTec AD 验证集 平均异常分数。**
语义分割模块
|模型名称|mloU（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |Deeplabv3_Plus-R50 |80.36|61.0531|1513.58|94.9 M|Deeplabv3_Plus-R50.yaml| |Deeplabv3_Plus-R101|81.10|100.026|2460.71|162.5 M|Deeplabv3_Plus-R101.yaml| |Deeplabv3-R50|79.90|82.2631|1735.83|138.3 M|Deeplabv3-R50.yaml| |Deeplabv3-R101|80.85|121.492|2685.51|205.9 M|Deeplabv3-R101.yaml| |OCRNet_HRNet-W18|80.67|48.2335|906.385|43.1 M|OCRNet_HRNet-W18.yaml| |OCRNet_HRNet-W48|82.15|78.9976|2226.95|249.8 M|OCRNet_HRNet-W48.yaml| |PP-LiteSeg-T|73.10|7.6827|138.683|28.5 M|PP-LiteSeg-T.yaml| |PP-LiteSeg-B|75.25|10.9935|194.727|47.0 M|PP-LiteSeg-B.yaml| |SegFormer-B0 (slice)|76.73|11.1946|268.929|13.2 M|SegFormer-B0.yaml| |SegFormer-B1 (slice)|78.35|17.9998|403.393|48.5 M|SegFormer-B1.yaml| |SegFormer-B2 (slice)|81.60|48.0371|1248.52|96.9 M|SegFormer-B2.yaml| |SegFormer-B3 (slice)|82.47|64.341|1666.35|167.3 M|SegFormer-B3.yaml| |SegFormer-B4 (slice)|82.38|82.4336|1995.42|226.7 M|SegFormer-B4.yaml| |SegFormer-B5 (slice)|82.58|97.3717|2420.19|229.7 M|SegFormer-B5.yaml|
**注：以上精度指标为 Cityscapes 数据集 mloU。**
|模型名称|mloU（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |SeaFormer_base(slice)|40.92|24.4073|397.574|30.8 M|SeaFormer_base.yaml| |SeaFormer_large (slice)|43.66|27.8123|550.464|49.8 M|SeaFormer_large.yaml| |SeaFormer_small (slice)|38.73|19.2295|358.343|14.3 M|SeaFormer_small.yaml| |SeaFormer_tiny (slice)|34.58|13.9496|330.132|6.1M |SeaFormer_tiny.yaml|
**注：以上精度指标为 ADE20k 数据集, slice 表示对输入图像进行了切图操作。**
实例分割模块
|模型名称|Mask AP|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |Mask-RT-DETR-H|50.6|132.693|4896.17|449.9 M|Mask-RT-DETR-H.yaml| |Mask-RT-DETR-L|45.7|46.5059|2575.92|113.6 M|Mask-RT-DETR-L.yaml| |Mask-RT-DETR-M|42.7|36.8329|-|66.6 M|Mask-RT-DETR-M.yaml| |Mask-RT-DETR-S|41.0|33.5007|-|51.8 M|Mask-RT-DETR-S.yaml| |Mask-RT-DETR-X|47.5|75.755|3358.04|237.5 M|Mask-RT-DETR-X.yaml| |Cascade-MaskRCNN-ResNet50-FPN|36.3|-|-|254.8 M|Cascade-MaskRCNN-ResNet50-FPN.yaml| |Cascade-MaskRCNN-ResNet50-vd-SSLDv2-FPN|39.1|-|-|254.7 M|Cascade-MaskRCNN-ResNet50-vd-SSLDv2-FPN.yaml| |MaskRCNN-ResNet50-FPN|35.6|-|-|157.5 M|MaskRCNN-ResNet50-FPN.yaml| |MaskRCNN-ResNet50-vd-FPN|36.4|-|-|157.5 M|MaskRCNN-ResNet50-vd-FPN.yaml| |MaskRCNN-ResNet50|32.8|-|-|127.8 M|MaskRCNN-ResNet50.yaml| |MaskRCNN-ResNet101-FPN|36.6|-|-|225.4 M|MaskRCNN-ResNet101-FPN.yaml| |MaskRCNN-ResNet101-vd-FPN|38.1|-|-|225.1 M|MaskRCNN-ResNet101-vd-FPN.yaml| |MaskRCNN-ResNeXt101-vd-FPN|39.5|-|-|370.0 M|MaskRCNN-ResNeXt101-vd-FPN.yaml| |PP-YOLOE_seg-S|32.5|-|-|31.5 M|PP-YOLOE_seg-S.yaml| |SOLOv2| 35.5|-|-|179.1 M|SOLOv2.yaml
**注：以上精度指标为 COCO2017 验证集 Mask AP(0.5:0.95)。**
文本检测模块
|模型名称|检测Hmean（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |PP-OCRv4_mobile_det |77.79|10.6923|120.177|4.2 M|PP-OCRv4_mobile_det.yaml| |PP-OCRv4_server_det |82.69|83.3501|2434.01|100.1M|PP-OCRv4_server_det.yaml|
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。
印章文本检测模块
|模型名称|检测Hmean（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |PP-OCRv4_mobile_seal_det|96.47|10.5878|131.813|4.7M |PP-OCRv4_mobile_seal_det.yaml| |PP-OCRv4_server_seal_det|98.21|84.341|2425.06|108.3 M|PP-OCRv4_server_seal_det.yaml|
注：以上精度指标的评估集是 PaddleX 自建的印章数据集，包含500印章图像。
文本识别模块
|模型名称|识别Avg Accuracy(%)|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |PP-OCRv4_mobile_rec |78.20|7.95018|46.7868|10.6 M|PP-OCRv4_mobile_rec.yaml| |PP-OCRv4_server_rec |79.20|7.19439|140.179|71.2 M|PP-OCRv4_server_rec.yaml|
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。
|模型名称|识别Avg Accuracy(%)|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |ch_SVTRv2_rec|68.81|8.36801|165.706|73.9 M|ch_SVTRv2_rec.yaml|
注：以上精度指标的评估集是 PaddleOCR算法模型挑战赛 - 赛题一：OCR端到端识别任务A榜。
|模型名称|识别Avg Accuracy(%)|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |ch_RepSVTR_rec|65.07|10.5047|51.5647|22.1 M|ch_RepSVTR_rec.yaml|
注：以上精度指标的评估集是 PaddleOCR算法模型挑战赛 - 赛题一：OCR端到端识别任务B榜。
公式识别模块
|模型名称|BLEU score|normed edit distance|ExpRate （%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-|-|-| |LaTeX_OCR_rec|0.8821|0.0823|40.01|-|-|89.7 M|LaTeX_OCR_rec.yaml|
注：以上精度指标测量自 LaTeX-OCR公式识别测试集。
表格结构识别模块
|模型名称|精度（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |SLANet|59.52|522.536|1845.37|6.9 M |SLANet.yaml| |SLANet_plus|63.69|522.536|1845.37|6.9 M |SLANet_plus.yaml|
**注：以上精度指标测量自 *PaddleX内部自建英文表格识别数据集*。**
图像矫正模块
|模型名称|MS-SSIM （%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |UVDoc|54.40|-|-|30.3 M|UVDoc.yaml|
**注：以上精度指标测量自 *PaddleX自建的图像矫正数据集*。**
版面区域检测模块
|模型名称|mAP@(0.50:0.95)（%）|GPU推理耗时（ms）|CPU推理耗时（ms）|模型存储大小|yaml 文件| |-|-|-|-|-|-| |PicoDet_layout_1x|86.8|13.036|91.2634|7.4 M |PicoDet_layout_1x.yaml| |PicoDet-S_layout_3cls|87.1|13.521 |45.7633|4.8 M|PicoDet-S_layout_3cls.yaml| |PicoDet-S_layout_17cls|70.3|13.5632|46.2059|4.8 M|PicoDet-S_layout_17cls.yaml| |PicoDet-L_layout_3cls|89.3|15.7425|159.771|22.6 M|PicoDet-L_layout_3cls.yaml| |PicoDet-L_layout_17cls|79.9|17.1901 |160.262|22.6 M|PicoDet-L_layout_17cls.yaml| |RT-DETR-H_layout_3cls|95.9|114.644|3832.62|470.1 M|RT-DETR-H_layout_3cls.yaml| |RT-DETR-H_layout_17cls|92.6|115.126|3827.25|470.2 M|RT-DETR-H_layout_17cls.yaml|
**注：以上精度指标的评估集是 *PaddleX 自建的版面区域检测数据集*，包含 1w 张图片。**
时序预测模块
|模型名称|mse|mae|模型存储大小|yaml 文件| |-|-|-|-|-| |DLinear|0.382|0.394|72 K|DLinear.yaml| |NLinear|0.386|0.392|40 K |NLinear.yaml| |Nonstationary|0.600|0.515|55.5 M|Nonstationary.yaml| |PatchTST|0.385|0.397|2.0 M |PatchTST.yaml| |RLinear|0.384|0.392|40 K|RLinear.yaml| |TiDE|0.405|0.412|31.7 M|TiDE.yaml| |TimesNet|0.417|0.431|4.9 M|TimesNet.yaml|
**注：以上精度指标测量自 ETTH1 数据集 *（在测试集test.csv上的评测结果）*。**
时序异常检测模块
|模型名称|precison|recall|f1_score|模型存储大小|yaml 文件| |-|-|-|-|-|-| |AutoEncoder_ad|99.36|84.36|91.25|52 K |AutoEncoder_ad.yaml| |DLinear_ad|98.98|93.96|96.41|112 K|DLinear_ad.yaml| |Nonstationary_ad|98.55|88.95|93.51|1.8 M |Nonstationary_ad.yaml| |PatchTST_ad|98.78|90.70|94.57|320 K |PatchTST_ad.yaml| |TimesNet_ad|98.37|94.80|96.56|1.3 M |TimesNet_ad.yaml|
**注：以上精度指标测量自 PSM 数据集。**
时序分类模块
|模型名称|acc(%)|模型存储大小|yaml 文件| |-|-|-|-| |TimesNet_cls|87.5|792 K|TimesNet_cls.yaml|
注：以上精度指标测量自 UWaveGestureLibrary数据集。
注：以上所有模型 GPU 推理耗时基于 NVIDIA Tesla T4 机器，精度类型为 FP32， CPU 推理速度基于 Intel(R) Xeon(R) Gold 5117 CPU @ 2.00GHz，线程数为8，精度类型为 FP32。