Skip to content

P
PaddleClas

PaddlePaddle / PaddleClas
接近 2 年 前同步成功

通知 116
Star 4999
Fork 1114
P
PaddleClas
提交 24168fb2 编写于 作者: weixin_46524038's avatar weixin_46524038 提交者： cuicheng01
浏览文件
操作

modify latency

上级 1b0658e6
展开全部 隐藏空白更改
内联 并排
Showing with 48 addition and 45 deletion
docs/zh_CN/models/ImageNet1k/GhostNet.md
浏览文件 @ 24168fb2
......@@ -73,7 +73,7 @@ GhostNet 是华为于 2020 年提出的一种全新的轻量化网络结构，
| GhostNet_x0_5 | 224 | 256 | 1.10 | 1.42 | 1.47 |
| GhostNet_x1_0 | 224 | 256 | 1.07 | 1.71 | 2.25 |
| GhostNet_x1_3 | 224 | 256 | 1.28 | 2.04 | 2.66 |
| GhostNet_x1_3_ssld | 224 | 256 | 1.85 | 3.17 | 4.29 |
| GhostNet_x1_3_ssld | 224 | 256 | 1.28 | 2.04 | 2.66 |
**备注：** 精度类型为 FP32，推理过程使用 TensorRT。
......
docs/zh_CN/models/ImageNet1k/HRNet.md
浏览文件 @ 24168fb2
......@@ -67,13 +67,13 @@ HRNet 是 2019 年由微软亚洲研究院提出的一种全新的神经网络
| Models | Size | Latency(ms)<br>bs=1 | Latency(ms)<br>bs=4 | Latency(ms)<br>bs=8 |
|-------------|-----------|-------------------|-------------------|-------------------|-------------------|
| HRNet_W18_C | 224 | 6.33 | 8.12 | 10.91 |
| HRNet_W18_C_ssld | 224 | 6.66 | 8.92 | 11.93 |
| HRNet_W18_C_ssld | 224 | 6.33 | 8.12 | 10.91 |
| HRNet_W30_C | 224 | 8.34 | 10.65 | 13.95 |
| HRNet_W32_C | 224 | 8.03 | 10.46 | 14.11 |
| HRNet_W40_C | 224 | 9.64 | 14.27 | 19.54 |
| HRNet_W44_C | 224 | 10.54 | 15.41 | 24.50 |
| HRNet_W48_C | 224 | 10.81 | 15.67 | 15.53 |
| HRNet_W48_C_ssld | 224 | 11.09 | 17.04 | 27.28 |
| HRNet_W48_C_ssld | 224 | 10.81 | 15.67 | 15.53 |
| HRNet_W64_C | 224 | 13.12 | 19.49 | 33.80 |
**备注：** 精度类型为 FP32，推理过程使用 TensorRT。
......
docs/zh_CN/models/ImageNet1k/NextViT.md
浏览文件 @ 24168fb2
......@@ -61,6 +61,9 @@ NextViT 是一种新的视觉 Transformer 网络，可以用作计算机视觉
| NextViT_small_224 | 224 | 7.76 | 10.86 | 14.20 |
| NextViT_base_224 | 224 | 12.02 | 16.21 | 20.63 |
| NextViT_large_224 | 224 | 16.51 | 21.91 | 27.25 |
| NextViT_small_224_ssld | 224 | 7.76 | 10.86 | 14.20 |
| NextViT_base_224_ssld | 224 | 12.02 | 16.21 | 20.63 |
| NextViT_large_224_ssld | 224 | 16.51 | 21.91 | 27.25 |
**备注：** 精度类型为 FP32，推理过程使用 TensorRT。
......
docs/zh_CN/models/ImageNet1k/PP-HGNet.md
浏览文件 @ 24168fb2
......@@ -75,20 +75,20 @@ PP-HGNet 与其他模型的比较如下，其中测试机器为 NVIDIA® Tesla®
| 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 |
| <b>PPHGNet_tiny<b> | <b>79.83<b> | <b>95.04<b> | <b>1.77<b> |
| <b>PPHGNet_tiny_ssld<b> | <b>81.95<b> | <b>96.12<b> | <b>1.77<b> |
| ResNet50 | 76.50 | 93.00 | 2.54 |
| ResNet50_vd | 79.12 | 94.44 | 2.60 |
| ResNet34 | 74.57 | 92.14 | 1.83 |
| ResNet34_vd | 75.98 | 92.98 | 1.87 |
| EfficientNetB0 | 77.38 | 93.31 | 1.58 |
| <b>PPHGNet_tiny<b> | <b>79.83<b> | <b>95.04<b> | <b>1.72<b> |
| <b>PPHGNet_tiny_ssld<b> | <b>81.95<b> | <b>96.12<b> | <b>1.72<b> |
| ResNet50 | 76.50 | 93.00 | 2.19 |
| ResNet50_vd | 79.12 | 94.44 | 2.23 |
| ResNet50_rsb | 80.40 | | 2.54 |
| EfficientNetB1 | 79.15 | 94.41 | 2.88 |
| EfficientNetB1 | 79.15 | 94.41 | 2.29 |
| SwinTransformer_tiny | 81.2 | 95.5 | 6.59 |
| <b>PPHGNet_small<b> | <b>81.51<b>| <b>95.82<b> | <b>2.52<b> |
| <b>PPHGNet_small_ssld<b> | <b>83.82<b>| <b>96.81<b> | <b>2.52<b> |
| Res2Net200_vd_26w_4s_ssld| 85.13 | 97.42 | 11.45 |
| ResNeXt101_32x48d_wsl | 85.37 | 97.69 | 55.07 |
| <b>PPHGNet_small<b> | <b>81.51<b>| <b>95.82<b> | <b>2.46<b> |
| <b>PPHGNet_small_ssld<b> | <b>83.82<b>| <b>96.81<b> | <b>2.46<b> |
| Res2Net200_vd_26w_4s_ssld| 85.13 | 97.42 | 10.80 |
| ResNeXt101_32x48d_wsl | 85.37 | 97.69 | 69.81 |
| SwinTransformer_base | 85.2 | 97.5 | 13.53 |
| <b>PPHGNet_base_ssld<b> | <b>85.00<b>| <b>97.35<b> | <b>5.97<b> |
......
docs/zh_CN/models/ImageNet1k/PP-LCNet.md
浏览文件 @ 24168fb2
......@@ -129,9 +129,9 @@ BaseNet 经过以上四个方面的改进，得到了 PP-LCNet。下表进一步
| PPLCNet_x1_5 | 4.5 | 342 | 73.71 | 91.53 | 0.54 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x1_5_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNet_x1_5_infer.tar) |
| PPLCNet_x2_0 | 6.5 | 590 | 75.18 | 92.27 | 0.64 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_0_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNet_x2_0_infer.tar) |
| PPLCNet_x2_5 | 9.0 | 906 | 76.60 | 93.00 | 0.71 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_5_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNet_x2_5_infer.tar) |
| PPLCNet_x0_5_ssld | 1.9 | 47 | 66.10 | 86.46 | 2.05 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_5_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNet_x0_5_ssld_infer.tar) |
| PPLCNet_x1_0_ssld | 3.0 | 161 | 74.39 | 92.09 | 2.46 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x1_0_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNet_x1_0_ssld_infer.tar) |
| PPLCNet_x2_5_ssld | 9.0 | 906 | 80.82 | 95.33 | 5.39 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_5_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNet_x2_5_ssld_infer.tar) |
| PPLCNet_x0_5_ssld | 1.9 | 47 | 66.10 | 86.46 | 0.44 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x0_5_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNet_x0_5_ssld_infer.tar) |
| PPLCNet_x1_0_ssld | 3.0 | 161 | 74.39 | 92.09 | 0.47 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x1_0_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNet_x1_0_ssld_infer.tar) |
| PPLCNet_x2_5_ssld | 9.0 | 906 | 80.82 | 95.33 | 0.71 | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNet_x2_5_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNet_x2_5_ssld_infer.tar) |
其中 `_ssld` 表示使用 `SSLD 蒸馏`后的模型。关于 `SSLD蒸馏` 的内容，详情 [SSLD 蒸馏](../../training/advanced/knowledge_distillation.md)。
......
docs/zh_CN/models/ImageNet1k/PP-LCNetV2.md
浏览文件 @ 24168fb2
......@@ -98,8 +98,8 @@ PPLCNetV2 目前提供的模型的精度、速度指标及预训练权重链接
| Model | Params(M) | FLOPs(M) | Top-1 Acc(\%) | Top-5 Acc(\%) | Latency(ms) | 预训练模型下载地址 | inference模型下载地址 |
|:--:|:--:|:--:|:--:|:--:|:--:|:--:|:--:|
| <b>PPLCNetV2_base<b> | <b>6.6<b> | <b>604<b> | <b>77.04<b> | <b>93.27<b> | <b>4.32<b> | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNetV2_base_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNetV2_base_infer.tar) |
| <b>PPLCNetV2_base_ssld<b> | <b>6.6<b> | <b>604<b> | <b>80.07<b> | <b>94.87<b> | <b>4.32<b> | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNetV2_base_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNetV2_base_ssld_infer.tar) |
| <b>PPLCNetV2_base<b> | <b>6.6<b> | <b>604<b> | <b>77.04<b> | <b>93.27<b> | <b>0.68<b> | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNetV2_base_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNetV2_base_infer.tar) |
| <b>PPLCNetV2_base_ssld<b> | <b>6.6<b> | <b>604<b> | <b>80.07<b> | <b>94.87<b> | <b>0.68<b> | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/legendary_models/PPLCNetV2_base_ssld_pretrained.pdparams) | [下载链接](https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/inference/PPLCNetV2_base_ssld_infer.tar) |
**备注：**
......@@ -113,7 +113,7 @@ PPLCNetV2 目前提供的模型的精度、速度指标及预训练权重链接
| MobileNetV3_Large_x1_25 | 7.4 | 714 | 76.4 | 93.00 | 5.19 |
| PPLCNetV1_x2_5 | 9 | 906 | 76.60 | 93.00 | 7.25 |
| <b>PPLCNetV2_base<b> | <b>6.6<b> | <b>604<b> | <b>77.04<b> | <b>93.27<b> | <b>0.68<b> |
| <b>PPLCNetV2_base_ssld<b> | <b>6.6<b> | <b>604<b> | <b>80.07<b> | <b>94.87<b> | <b>4.32<b> |
| <b>PPLCNetV2_base_ssld<b> | <b>6.6<b> | <b>604<b> | <b>80.07<b> | <b>94.87<b> | <b>0.68<b> |
<a name="2"></a>
......
docs/zh_CN/models/ImageNet1k/Res2Net.md
浏览文件 @ 24168fb2
......@@ -69,8 +69,8 @@ Res2Net 是 2019 年提出的一种全新的对 ResNet 的改进方案，该方
| Res2Net50_14w_8s | 224 | 4.13 | 6.56 | 9.45 |
| Res2Net101_vd_26w_4s | 224 | 5.96 | 10.56 | 15.20 |
| Res2Net200_vd_26w_4s | 224 | 10.80 | 19.48 | 27.95 |
| Res2Net50_vd_26w_4s_ssld | 224 | 3.58 | 6.35 | 9.52 |
| Res2Net101_vd_26w_4s_ssld | 224 | 9.56 | 10.56 | 15.20 |
| Res2Net50_vd_26w_4s_ssld | 224 | 3.35 | 5.79 | 8.63 |
| Res2Net101_vd_26w_4s_ssld | 224 | 5.96 | 10.56 | 15.20 |
| Res2Net200_vd_26w_4s_ssld | 224 | 10.80 | 19.48 | 27.95 |
**备注：** 精度类型为 FP32，推理过程使用 TensorRT。
......
docs/zh_CN/models/ImageNet1k/ResNet.md
浏览文件 @ 24168fb2
......@@ -87,7 +87,7 @@ PaddleClas 提供的 ResNet 系列的模型包括 ResNet50，ResNet50_vd，ResNe
| ResNet18_vd | 224 | 1.11 | 1.52 | 2.60 |
| ResNet34 | 224 | 1.83 | 2.41 | 4.23 |
| ResNet34_vd | 224 | 1.87 | 2.49 | 4.41 |
| ResNet34_vd_ssld | 224 | 2.00 | 3.26 | 5.85 |
| ResNet34_vd_ssld | 224 | 1.87 | 2.49 | 4.41 |
| ResNet50 | 224 | 2.19 | 3.77 | 6.22 |
| ResNet50_vc | 224 | 2.57 | 4.83 | 7.52 |
| ResNet50_vd | 224 | 2.23 | 3.92 | 6.46 |
......@@ -99,8 +99,8 @@ PaddleClas 提供的 ResNet 系列的模型包括 ResNet50，ResNet50_vd，ResNe
| SE_ResNet18_vd | 224 | 1.31 | 1.77 | 2.92 |
| SE_ResNet34_vd | 224 | 2.20 | 2.99 | 5.09 |
| SE_ResNet50_vd | 224 | 2.72 | 5.07 | 8.12 |
| ResNet50_vd_ssld | 224 | 2.59 | 4.87 | 7.62 |
| ResNet101_vd_ssld | 224 | 4.43 | 8.25 | 12.58 |
| ResNet50_vd_ssld | 224 | 2.23 | 3.92 | 6.46 |
| ResNet101_vd_ssld | 224 | 4.04 | 6.84 | 11.44 |
**备注：** 精度类型为 FP32，推理过程使用 TensorRT。
......
Markdown is supported
0% 或 .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑！
想要评论请 注册 或