Update pretrain model.

1. Add training log and pretrain
deepinsight · Mar 12, 2021 · f89ecaa · f89ecaa
1 parent d873824
commit f89ecaa
Show file tree

Hide file tree

Showing 2 changed files with 23 additions and 33 deletions.
diff --git a/recognition/arcface_torch/README.md b/recognition/arcface_torch/README.md
@@ -9,6 +9,10 @@ torch >= 1.6.0
 ```
 
 ## Train on a single node 
+If you want to use 8 GPU to train, you should set `--nproc_per_node=8` and set `CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 `  
+If you want to use 4 GPU to train, you should set `--nproc_per_node=4` and set `CUDA_VISIBLE_DEVICES=0,1,2,3`  
+If you want to use 1 GPU to train, you should set `--nproc_per_node=1` ...  
+
 ```shell
 export OMP_NUM_THREADS=4
 export CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 
@@ -42,39 +46,26 @@ CUDA_VISIBLE_DEVICES=0,1 python eval_ijbc.py \
 More details see [eval.md](docs/eval.md) in docs.
 
 
-## Model Zoo
-### MS1MV3
-|   Datasets          | backbone | IJBC(1e-05) | IJBC(1e-04) |agedb30|cfp_fp|lfw  | 
-| :---:               | :---     | :---        | :---        |:---   |:---  |:--- |  
-| MS1MV3-Arcface      | r18      |   92.08     |  94.68      |97.65  |97.63 |99.73|
-| MS1MV3-Arcface      | r34      |             |             |       |      |     | 
-| MS1MV3-Arcface      | r50      |   94.79     |  96.43      |98.28  |98.89 |99.85| 
-| MS1MV3-Arcface      | r50-amp  |   94.72     |  96.41      |98.30  |99.06 |99.85| 
-| MS1MV3-Arcface      | r100     |   95.22     |  96.87      |98.45  |99.19 |99.85| 
-
-### Glint360k
-|   Datasets          | backbone | IJBC(1e-05) | IJBC(1e-04) |agedb30|cfp_fp|lfw  | 
-| :---:               | :---     | :---        | :---        |:---   |:---  |:--- |
-| Glint360k-Cosface   | r100     |  96.19      | 97.39       |98.52  |99.26 |99.83|
+## Model Zoo  
 
-More details see [eval.md](docs/modelzoo.md) in docs.
+The models are available for non-commercial research purposes only.
 
+All Model Can be found in here.  
+[Baidu Yun Pan](https://pan.baidu.com/s/1CL-l4zWqsI1oDuEEYVhj-g):   e8pw  
 
+### MS1MV3
+|   Datasets          |    log     | backbone    | IJBC(1e-05) | IJBC(1e-04) |agedb30|cfp_fp|lfw  | 
+| :---:               |    :---    | :---        | :---        | :---        |:---   |:---  |:--- |  
+| MS1MV3-Arcface      |[log](https://raw.githubusercontent.com/anxiangsir/insightface_arcface_log/master/ms1mv3_arcface_r18/training.log)             | r18                 | 92.08 | 94.68 | 97.65 | 97.63 | 99.73 |
+| MS1MV3-Arcface      |[log](https://raw.githubusercontent.com/anxiangsir/insightface_arcface_log/master/ms1mv3_arcface_r34/training.log)             | r34                 | 94.13 | 95.98 | 98.05 | 98.60 | 99.80 | 
+| MS1MV3-Arcface      |[log](https://raw.githubusercontent.com/anxiangsir/insightface_arcface_log/master/ms1mv3_arcface_r50/training.log)             | r50                 | 94.79 | 96.43 | 98.28 | 98.89 | 99.85 | 
+| MS1MV3-Arcface      |[log](https://raw.githubusercontent.com/anxiangsir/insightface_arcface_log/master/ms1mv3_arcface_r50_fp16/training.log)        | r50-fp16            | 94.72 | 96.41 | 98.30 | 99.06 | 99.85 | 
+| MS1MV3-Arcface      |[log](https://raw.githubusercontent.com/anxiangsir/insightface_arcface_log/master/ms1mv3_arcface_r100/training.log)            | r100                | 95.22 | 96.87 | 98.45 | 99.19 | 99.85 | 
+
+### Glint360k
+|   Datasets          | log   |backbone               | IJBC(1e-05) | IJBC(1e-04) |agedb30|cfp_fp|lfw  | 
+| :---:               | :---  |:---                   | :---        | :---        |:---   |:---  |:--- |
+| Glint360k-Cosface   |[log](https://raw.githubusercontent.com/anxiangsir/insightface_arcface_log/master/glint360k_cosface_r100/training.log)         |r100                 | 96.19 | 97.39 | 98.52 | 99.26 | 99.83 |
+| Glint360k-Cosface   |[log](https://raw.githubusercontent.com/anxiangsir/insightface_arcface_log/master/glint360k_cosface_r100_fp16_0.1/training.log)|r100-fp16-sample-0.1 | 95.95 | 97.35 | 98.57 | 99.30 | 99.85 |
 
-## Citation
-```
-@inproceedings{deng2019arcface,
-  title={Arcface: Additive angular margin loss for deep face recognition},
-  author={Deng, Jiankang and Guo, Jia and Xue, Niannan and Zafeiriou, Stefanos},
-  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
-  pages={4690--4699},
-  year={2019}
-}
-@inproceedings{an2020partical_fc,
-  title={Partial FC: Training 10 Million Identities on a Single Machine},
-  author={An, Xiang and Zhu, Xuhan and Xiao, Yang and Wu, Lan and Zhang, Ming and Gao, Yuan and Qin, Bin and
-  Zhang, Debing and Fu Ying},
-  booktitle={Arxiv 2010.05222},
-  year={2020}
-}
-```
+More details see [eval.md](docs/modelzoo.md) in docs.
diff --git a/recognition/arcface_torch/backbones/iresnet.py b/recognition/arcface_torch/backbones/iresnet.py
@@ -27,7 +27,6 @@ def conv1x1(in_planes, out_planes, stride=1):
 
 class IBasicBlock(nn.Module):
     expansion = 1
-
     def __init__(self, inplanes, planes, stride=1, downsample=None,
                  groups=1, base_width=64, dilation=1):
         super(IBasicBlock, self).__init__()