提交 e3324859 编写于 作者: littletomatodonkey's avatar littletomatodonkey

add res2net200_vd_26w_4s_ssld doc

上级 ad6b6deb
...@@ -82,6 +82,7 @@ Training process is carried out on the large-scale dataset with 5 million images ...@@ -82,6 +82,7 @@ Training process is carried out on the large-scale dataset with 5 million images
| MobileNetV3_small_x1_0 | 360 | 1e-5 | 5760/24 | 3.65625 | cosine_decay_warmup | 70.11% | | MobileNetV3_small_x1_0 | 360 | 1e-5 | 5760/24 | 3.65625 | cosine_decay_warmup | 70.11% |
| ResNet50_vd | 360 | 7e-5 | 1024/32 | 0.4 | cosine_decay_warmup | 82.07% | | ResNet50_vd | 360 | 7e-5 | 1024/32 | 0.4 | cosine_decay_warmup | 82.07% |
| ResNet101_vd | 360 | 7e-5 | 1024/32 | 0.4 | cosine_decay_warmup | 83.41% | | ResNet101_vd | 360 | 7e-5 | 1024/32 | 0.4 | cosine_decay_warmup | 83.41% |
| Res2Net200_vd_26w_4s | 360 | 4e-5 | 1024/32 | 0.4 | cosine_decay_warmup | 84.82% |
## finetuning using ImageNet1k ## finetuning using ImageNet1k
...@@ -96,6 +97,7 @@ Finetuning is carried out on ImageNet1k dataset to restore distribution between ...@@ -96,6 +97,7 @@ Finetuning is carried out on ImageNet1k dataset to restore distribution between
| MobileNetV3_small_x1_0 | 30 | 1e-5 | 6400/32 | 0.025 | cosine_decay_warmup | 71.28% | | MobileNetV3_small_x1_0 | 30 | 1e-5 | 6400/32 | 0.025 | cosine_decay_warmup | 71.28% |
| ResNet50_vd | 60 | 7e-5 | 1024/32 | 0.004 | cosine_decay_warmup | 82.39% | | ResNet50_vd | 60 | 7e-5 | 1024/32 | 0.004 | cosine_decay_warmup | 82.39% |
| ResNet101_vd | 30 | 7e-5 | 1024/32 | 0.004 | cosine_decay_warmup | 83.73% | | ResNet101_vd | 30 | 7e-5 | 1024/32 | 0.004 | cosine_decay_warmup | 83.73% |
| Res2Net200_vd_26w_4s | 360 | 4e-5 | 1024/32 | 0.004 | cosine_decay_warmup | 85.13% |
## Data agmentation and Fix strategy ## Data agmentation and Fix strategy
......
...@@ -33,6 +33,7 @@ At present, there are a total of 24 pretrained models of the three categories op ...@@ -33,6 +33,7 @@ At present, there are a total of 24 pretrained models of the three categories op
| Res2Net50_14w_8s | 0.795 | 0.947 | 0.781 | 0.939 | 9.010 | 25.720 | | Res2Net50_14w_8s | 0.795 | 0.947 | 0.781 | 0.939 | 9.010 | 25.720 |
| Res2Net101_vd_26w_4s | 0.806 | 0.952 | | | 16.670 | 45.220 | | Res2Net101_vd_26w_4s | 0.806 | 0.952 | | | 16.670 | 45.220 |
| Res2Net200_vd_26w_4s | 0.812 | 0.957 | | | 31.490 | 76.210 | | Res2Net200_vd_26w_4s | 0.812 | 0.957 | | | 31.490 | 76.210 |
| Res2Net200_vd_26w_4s_ssld | **0.851** | 0.974 | | | 31.490 | 76.210 |
| ResNeXt50_32x4d | 0.778 | 0.938 | 0.778 | | 8.020 | 23.640 | | ResNeXt50_32x4d | 0.778 | 0.938 | 0.778 | | 8.020 | 23.640 |
| ResNeXt50_vd_32x4d | 0.796 | 0.946 | | | 8.500 | 23.660 | | ResNeXt50_vd_32x4d | 0.796 | 0.946 | | | 8.500 | 23.660 |
| ResNeXt50_64x4d | 0.784 | 0.941 | | | 15.060 | 42.360 | | ResNeXt50_64x4d | 0.784 | 0.941 | | | 15.060 | 42.360 |
......
...@@ -130,6 +130,7 @@ python tools/infer/predict.py \ ...@@ -130,6 +130,7 @@ python tools/infer/predict.py \
- [Res2Net50_14w_8s](https://paddle-imagenet-models-name.bj.bcebos.com/Res2Net50_14w_8s_pretrained.tar) - [Res2Net50_14w_8s](https://paddle-imagenet-models-name.bj.bcebos.com/Res2Net50_14w_8s_pretrained.tar)
- [Res2Net101_vd_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/Res2Net101_vd_26w_4s_pretrained.tar) - [Res2Net101_vd_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/Res2Net101_vd_26w_4s_pretrained.tar)
- [Res2Net200_vd_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/Res2Net200_vd_26w_4s_pretrained.tar) - [Res2Net200_vd_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/Res2Net200_vd_26w_4s_pretrained.tar)
- [Res2Net200_vd_26w_4s_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/Res2Net200_vd_26w_4s_ssld_pretrained.tar)
- Inception series - Inception series
......
# Release Notes # Release Notes
* 2020.07.14
* Add `Res2Net200_vd_26w_4s_ssld` pretrained model, whose Top-1 Acc on ImageNet1k dataset reaches 85.13%.
* Add `Fix_ResNet50_vd_ssld_v2` pretrained model, whose Top-1 Acc on ImageNet1k dataset reaches 84.00%.
* 2020.06.17 * 2020.06.17
* Add English documents。 * Add English documents。
......
...@@ -85,6 +85,7 @@ SSLD的流程图如下图所示。 ...@@ -85,6 +85,7 @@ SSLD的流程图如下图所示。
| MobileNetV3_small_x1_0 | 360 | 1e-5 | 5760/24 | 3.65625 | cosine_decay_warmup | 70.11% | | MobileNetV3_small_x1_0 | 360 | 1e-5 | 5760/24 | 3.65625 | cosine_decay_warmup | 70.11% |
| ResNet50_vd | 360 | 7e-5 | 1024/32 | 0.4 | cosine_decay_warmup | 82.07% | | ResNet50_vd | 360 | 7e-5 | 1024/32 | 0.4 | cosine_decay_warmup | 82.07% |
| ResNet101_vd | 360 | 7e-5 | 1024/32 | 0.4 | cosine_decay_warmup | 83.41% | | ResNet101_vd | 360 | 7e-5 | 1024/32 | 0.4 | cosine_decay_warmup | 83.41% |
| Res2Net200_vd_26w_4s | 360 | 4e-5 | 1024/32 | 0.4 | cosine_decay_warmup | 84.82% |
## 3.3 ImageNet1k训练集finetune ## 3.3 ImageNet1k训练集finetune
...@@ -99,6 +100,7 @@ SSLD的流程图如下图所示。 ...@@ -99,6 +100,7 @@ SSLD的流程图如下图所示。
| MobileNetV3_small_x1_0 | 30 | 1e-5 | 6400/32 | 0.025 | cosine_decay_warmup | 71.28% | | MobileNetV3_small_x1_0 | 30 | 1e-5 | 6400/32 | 0.025 | cosine_decay_warmup | 71.28% |
| ResNet50_vd | 60 | 7e-5 | 1024/32 | 0.004 | cosine_decay_warmup | 82.39% | | ResNet50_vd | 60 | 7e-5 | 1024/32 | 0.004 | cosine_decay_warmup | 82.39% |
| ResNet101_vd | 30 | 7e-5 | 1024/32 | 0.004 | cosine_decay_warmup | 83.73% | | ResNet101_vd | 30 | 7e-5 | 1024/32 | 0.004 | cosine_decay_warmup | 83.73% |
| Res2Net200_vd_26w_4s | 360 | 4e-5 | 1024/32 | 0.004 | cosine_decay_warmup | 85.13% |
## 3.4 数据增广以及基于Fix策略的微调 ## 3.4 数据增广以及基于Fix策略的微调
......
...@@ -32,6 +32,7 @@ Res2Net是2019年提出的一种全新的对ResNet的改进方案,该方案可 ...@@ -32,6 +32,7 @@ Res2Net是2019年提出的一种全新的对ResNet的改进方案,该方案可
| Res2Net50_14w_8s | 0.795 | 0.947 | 0.781 | 0.939 | 9.010 | 25.720 | | Res2Net50_14w_8s | 0.795 | 0.947 | 0.781 | 0.939 | 9.010 | 25.720 |
| Res2Net101_vd_26w_4s | 0.806 | 0.952 | | | 16.670 | 45.220 | | Res2Net101_vd_26w_4s | 0.806 | 0.952 | | | 16.670 | 45.220 |
| Res2Net200_vd_26w_4s | 0.812 | 0.957 | | | 31.490 | 76.210 | | Res2Net200_vd_26w_4s | 0.812 | 0.957 | | | 31.490 | 76.210 |
| Res2Net200_vd_26w_4s | **0.851** | 0.974 | | | 31.490 | 76.210 |
| ResNeXt50_32x4d | 0.778 | 0.938 | 0.778 | | 8.020 | 23.640 | | ResNeXt50_32x4d | 0.778 | 0.938 | 0.778 | | 8.020 | 23.640 |
| ResNeXt50_vd_32x4d | 0.796 | 0.946 | | | 8.500 | 23.660 | | ResNeXt50_vd_32x4d | 0.796 | 0.946 | | | 8.500 | 23.660 |
| ResNeXt50_64x4d | 0.784 | 0.941 | | | 15.060 | 42.360 | | ResNeXt50_64x4d | 0.784 | 0.941 | | | 15.060 | 42.360 |
......
...@@ -130,6 +130,7 @@ python tools/infer/predict.py \ ...@@ -130,6 +130,7 @@ python tools/infer/predict.py \
- [Res2Net50_14w_8s](https://paddle-imagenet-models-name.bj.bcebos.com/Res2Net50_14w_8s_pretrained.tar) - [Res2Net50_14w_8s](https://paddle-imagenet-models-name.bj.bcebos.com/Res2Net50_14w_8s_pretrained.tar)
- [Res2Net101_vd_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/Res2Net101_vd_26w_4s_pretrained.tar) - [Res2Net101_vd_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/Res2Net101_vd_26w_4s_pretrained.tar)
- [Res2Net200_vd_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/Res2Net200_vd_26w_4s_pretrained.tar) - [Res2Net200_vd_26w_4s](https://paddle-imagenet-models-name.bj.bcebos.com/Res2Net200_vd_26w_4s_pretrained.tar)
- [Res2Net200_vd_26w_4s_ssld](https://paddle-imagenet-models-name.bj.bcebos.com/Res2Net200_vd_26w_4s_ssld_pretrained.tar)
- Inception系列 - Inception系列
......
# 更新日志 # 更新日志
* 2020.07.14
* 添加Res2Net200_vd_26w_4s_ssld模型,在ImageNet上Top-1 Acc可达85.13%。
* 添加Fix_ResNet50_vd_ssld_v2模型,,在ImageNet上Top-1 Acc可达84.0%。
* 2020.06.17 * 2020.06.17
* 添加英文文档。 * 添加英文文档。
......
Markdown is supported
0% .
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