update docs

thuml · Oct 16, 2021 · 1cd7e04 · 1cd7e04
1 parent d527e4e
commit 1cd7e04
Show file tree

Hide file tree

Showing 15 changed files with 107 additions and 118 deletions.
diff --git a/common/vision/datasets/_util.py b/common/vision/datasets/_util.py
@@ -29,8 +29,8 @@ def download(root: str, file_name: str, archive_name: str, url_link: str):
             download_and_extract_archive(url_link, download_root=root, filename=archive_name, remove_finished=False)
         except Exception:
             print("Fail to download {} from url link {}".format(archive_name, url_link))
-            print('Please check you internet connection or '
-                  "reinstall DALIB by 'pip install --upgrade dalib'")
+            print('Please check you internet connection.'
+                  "Simply trying again may be fine.")
             exit(0)
 
 

diff --git a/docs/dalib/benchmarks/image_classification.rst b/docs/dalib/benchmarks/image_classification.rst
@@ -136,7 +136,7 @@ DomainNet accuracy on ResNet-101
 Methods     c->p    c->r    c->s    p->c    p->r    p->s    r->c    r->p    r->s    s->c    s->p    s->r    Avg
 Source Only 32.7    50.6    39.4    41.1    56.8    35.0    48.6    48.8    36.1    49.0    34.8    46.1    43.3
 DANN        37.9    54.3    44.4    41.7    55.6    36.8    50.7    50.8    40.1    55.0    45.0    54.5    47.2
-ADDA
+ADDA        38.4    54.1    44.1    43.5    56.7    39.2    52.8    51.3    40.9    55.0    45.4    54.5    48.0
 ADDAgrl
 DAN         38.8    55.2    43.9    45.9    59.0    40.8    50.8    49.8    38.9    56.1    45.9    55.5    48.4
 JAN         40.5    56.7    45.1    47.2    59.9    43.0    54.2    52.6    41.9    56.6    46.2    55.5    50.0

diff --git a/docs/dalib/benchmarks/image_regression.rst b/docs/dalib/benchmarks/image_regression.rst
@@ -2,7 +2,7 @@
 Image Regression
 ===============================================
 
-We provide benchmarks of different domain adaptation algorithms on `dSprites`_.
+We provide benchmarks of different domain adaptation algorithms on `dSprites`_ and `MPI3D`_ .
 Those domain adaptation algorithms includes:
 
 -  :ref:`MDD`
@@ -28,8 +28,16 @@ dSprites error on ResNet-18
 ===========     ======  ======  ======  ======  ======  ======  ======
 Methods         Avg     C → N   C → S   N → C   N → S   S → C   S → N
 Source Only     0.157   0.232   0.271   0.081   0.220   0.038   0.092
-DD              0.057   0.047	0.080	0.030	0.095	0.053	0.037
+DD              0.057   0.047   0.080   0.030   0.095   0.053   0.037
 ===========     ======  ======  ======  ======  ======  ======  ======
 
 
+.. _MPI3D:
 
+MPI3D error on ResNet-18
+---------------------------------
+===========     ======  ======  ======  ======  ======  ======  ======
+Methods         Avg     RL → RC RL → T  RC → RL RC → T  T → RL  T → RC
+Source Only     0.176   0.232   0.271   0.081   0.220   0.038   0.092
+DD              0.030   0.086   0.029   0.057   0.189   0.131   0.087
+===========     ======  ======  ======  ======  ======  ======  ======
diff --git a/docs/dglib/benchmarks/classification.rst → ...dglib/benchmarks/image_classification.rst b/docs/dglib/benchmarks/classification.rst → ...dglib/benchmarks/image_classification.rst
@@ -2,7 +2,7 @@
 Image Classification
 ===============================
 
-We provide benchmarks of different domain generalization algorithms on `PACS`_, `Office-Home`_, `DomainNet`_,
+We provide benchmarks of different domain generalization algorithms on `PACS`_, `Office-Home`_,
 `iWildCam-Wilds`_, `Camelyon17-Wilds`_, `FMoW-Wilds`_.
 Those domain generalization algorithms includes:
 
@@ -69,24 +69,6 @@ GroupDRO 70.0  66.7  55.2  78.8  79.9
 CORAL    70.9  68.3  55.4  78.8  81.0
 ======== ===== ===== ===== ===== =====
 
-.. _DomainNet:
-
------------------------------------
-DomainNet accuracy on ResNet-50
------------------------------------
-
-======== ===== ========= =========== ========== =========== ====== ========
-Methods   avg   clipart   infograph   painting   quickdraw   real   sketch
-ERM
-IBN
-MixStyle
-MLDG
-IRM
-VREx
-GroupDRO
-CORAL
-======== ===== ========= =========== ========== =========== ====== ========
-
 .. _iWildCam-Wilds:
 
 -----------------------------------

diff --git a/docs/dglib/benchmarks/reid.rst → docs/dglib/benchmarks/re_identification.rst b/docs/dglib/benchmarks/reid.rst → docs/dglib/benchmarks/re_identification.rst
@@ -1,5 +1,5 @@
 ===============================
-Person Re-Identification
+Re-Identification
 ===============================
 
 We provide benchmarks of different domain generalization algorithms. Currently three datasets are supported:

diff --git a/docs/index.rst b/docs/index.rst
@@ -41,8 +41,8 @@ Transfer Learning
     :caption: Domain Generalization Settings
     :titlesonly:
 
-    dglib/benchmarks/classification
-    dglib/benchmarks/reid
+    dglib/benchmarks/image_classification
+    dglib/benchmarks/re_identification
 
 
 .. toctree::

diff --git a/examples/domain_adaptation/image_regression/dann.sh b/examples/domain_adaptation/image_regression/dann.sh
@@ -1,5 +1,5 @@
 # DSprites
-CUDA_VISIBLE_DEVICES=2 python dann.py data/dSprites -d DSprites -s C -t N -a resnet18 --epochs 40 --seed 0 --log logs/dann/DSprites_C2N
+CUDA_VISIBLE_DEVICES=0 python dann.py data/dSprites -d DSprites -s C -t N -a resnet18 --epochs 40 --seed 0 --log logs/dann/DSprites_C2N
 CUDA_VISIBLE_DEVICES=0 python dann.py data/dSprites -d DSprites -s C -t S -a resnet18 --epochs 40 --seed 0 --log logs/dann/DSprites_C2S
 CUDA_VISIBLE_DEVICES=0 python dann.py data/dSprites -d DSprites -s N -t C -a resnet18 --epochs 40 --seed 0 --log logs/dann/DSprites_N2C
 CUDA_VISIBLE_DEVICES=0 python dann.py data/dSprites -d DSprites -s N -t S -a resnet18 --epochs 40 --seed 0 --log logs/dann/DSprites_N2S

diff --git a/examples/domain_adaptation/image_regression/dd.sh b/examples/domain_adaptation/image_regression/dd.sh
@@ -1,15 +1,15 @@
 # DSprites
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/dSprites -d DSprites -s C -t N -a resnet18 --epochs 40 --seed 0 -b 128 --log logs/mdd/dSprites_C2N --wd 0.0005
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/dSprites -d DSprites -s C -t S -a resnet18 --epochs 40 --seed 0 -b 128 --log logs/mdd/dSprites_C2S --wd 0.0005
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/dSprites -d DSprites -s N -t C -a resnet18 --epochs 40 --seed 0 -b 128 --log logs/mdd/dSprites_N2C --wd 0.0005
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/dSprites -d DSprites -s N -t S -a resnet18 --epochs 40 --seed 0 -b 128 --log logs/mdd/dSprites_N2S --wd 0.0005
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/dSprites -d DSprites -s S -t C -a resnet18 --epochs 40 --seed 0 -b 128 --log logs/mdd/dSprites_S2C --wd 0.0005
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/dSprites -d DSprites -s S -t N -a resnet18 --epochs 40 --seed 0 -b 128 --log logs/mdd/dSprites_S2N --wd 0.0005
+CUDA_VISIBLE_DEVICES=0 python dd.py data/dSprites -d DSprites -s C -t N -a resnet18 --epochs 40 --seed 0 -b 128 --log logs/dd/dSprites_C2N --wd 0.0005
+CUDA_VISIBLE_DEVICES=0 python dd.py data/dSprites -d DSprites -s C -t S -a resnet18 --epochs 40 --seed 0 -b 128 --log logs/dd/dSprites_C2S --wd 0.0005
+CUDA_VISIBLE_DEVICES=0 python dd.py data/dSprites -d DSprites -s N -t C -a resnet18 --epochs 40 --seed 0 -b 128 --log logs/dd/dSprites_N2C --wd 0.0005
+CUDA_VISIBLE_DEVICES=0 python dd.py data/dSprites -d DSprites -s N -t S -a resnet18 --epochs 40 --seed 0 -b 128 --log logs/dd/dSprites_N2S --wd 0.0005
+CUDA_VISIBLE_DEVICES=0 python dd.py data/dSprites -d DSprites -s S -t C -a resnet18 --epochs 40 --seed 0 -b 128 --log logs/dd/dSprites_S2C --wd 0.0005
+CUDA_VISIBLE_DEVICES=0 python dd.py data/dSprites -d DSprites -s S -t N -a resnet18 --epochs 40 --seed 0 -b 128 --log logs/dd/dSprites_S2N --wd 0.0005
 
 # MPI3D
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/mpi3d -d MPI3D -s RL -t RC -a resnet18 --epochs 60 --seed 0 -b 36 --log logs/mdd/MPI3D_RL2RC --normalization IN --resize-size 224 --weight-decay 0.001
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/mpi3d -d MPI3D -s RL -t T -a resnet18 --epochs 60 --seed 0 -b 36 --log logs/mdd/MPI3D_RL2T --normalization IN --resize-size 224 --weight-decay 0.001
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/mpi3d -d MPI3D -s RC -t RL -a resnet18 --epochs 60 --seed 0 -b 36 --log logs/mdd/MPI3D_RC2RL --normalization IN --resize-size 224 --weight-decay 0.001
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/mpi3d -d MPI3D -s RC -t T -a resnet18 --epochs 60 --seed 0 -b 36 --log logs/mdd/MPI3D_RC2T --normalization IN --resize-size 224 --weight-decay 0.001
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/mpi3d -d MPI3D -s T -t RL -a resnet18 --epochs 60 --seed 0 -b 36 --log logs/mdd/MPI3D_T2RL --normalization IN --resize-size 224 --weight-decay 0.001
-CUDA_VISIBLE_DEVICES=0 python mdd.py data/mpi3d -d MPI3D -s T -t RC -a resnet18 --epochs 60 --seed 0 -b 36 --log logs/mdd/MPI3D_T2RC --normalization IN --resize-size 224 --weight-decay 0.001
+CUDA_VISIBLE_DEVICES=0 python dd.py data/mpi3d -d MPI3D -s RL -t RC -a resnet18 --epochs 60 --seed 0 -b 36 --log logs/dd/MPI3D_RL2RC --normalization IN --resize-size 224 --weight-decay 0.001
+CUDA_VISIBLE_DEVICES=0 python dd.py data/mpi3d -d MPI3D -s RL -t T -a resnet18 --epochs 60 --seed 0 -b 36 --log logs/dd/MPI3D_RL2T --normalization IN --resize-size 224 --weight-decay 0.001
+CUDA_VISIBLE_DEVICES=0 python dd.py data/mpi3d -d MPI3D -s RC -t RL -a resnet18 --epochs 60 --seed 0 -b 36 --log logs/dd/MPI3D_RC2RL --normalization IN --resize-size 224 --weight-decay 0.001
+CUDA_VISIBLE_DEVICES=0 python dd.py data/mpi3d -d MPI3D -s RC -t T -a resnet18 --epochs 60 --seed 0 -b 36 --log logs/dd/MPI3D_RC2T --normalization IN --resize-size 224 --weight-decay 0.001
+CUDA_VISIBLE_DEVICES=0 python dd.py data/mpi3d -d MPI3D -s T -t RL -a resnet18 --epochs 60 --seed 0 -b 36 --log logs/dd/MPI3D_T2RL --normalization IN --resize-size 224 --weight-decay 0.001
+CUDA_VISIBLE_DEVICES=0 python dd.py data/mpi3d -d MPI3D -s T -t RC -a resnet18 --epochs 60 --seed 0 -b 36 --log logs/dd/MPI3D_T2RC --normalization IN --resize-size 224 --weight-decay 0.001
diff --git a/examples/domain_adaptation/image_regression/rsd.sh b/examples/domain_adaptation/image_regression/rsd.sh
@@ -1,5 +1,5 @@
 # DSprites
-CUDA_VISIBLE_DEVICES=1 python rsd.py data/dSprites -d DSprites -s C -t N -a resnet18 --epochs 40 --seed 0 --log logs/rsd/DSprites_C2N
+CUDA_VISIBLE_DEVICES=0 python rsd.py data/dSprites -d DSprites -s C -t N -a resnet18 --epochs 40 --seed 0 --log logs/rsd/DSprites_C2N
 CUDA_VISIBLE_DEVICES=0 python rsd.py data/dSprites -d DSprites -s C -t S -a resnet18 --epochs 40 --seed 0 --log logs/rsd/DSprites_C2S
 CUDA_VISIBLE_DEVICES=0 python rsd.py data/dSprites -d DSprites -s N -t C -a resnet18 --epochs 40 --seed 0 --log logs/rsd/DSprites_N2C
 CUDA_VISIBLE_DEVICES=0 python rsd.py data/dSprites -d DSprites -s N -t S -a resnet18 --epochs 40 --seed 0 --log logs/rsd/DSprites_N2S

diff --git a/examples/domain_adaptation/re_identification/README.md b/examples/domain_adaptation/re_identification/README.md
@@ -55,18 +55,17 @@ If you use these methods in your research, please consider citing.
 
 ```
 @inproceedings{IBN-Net,  
-  author = {Xingang Pan, Ping Luo, Jianping Shi, and Xiaoou Tang},  
-  title = {Two at Once: Enhancing Learning and Generalization Capacities via IBN-Net},  
-  booktitle = {ECCV},  
-  year = {2018}  
+    author = {Xingang Pan, Ping Luo, Jianping Shi, and Xiaoou Tang},  
+    title = {Two at Once: Enhancing Learning and Generalization Capacities via IBN-Net},  
+    booktitle = {ECCV},  
+    year = {2018}  
 }
 
 @inproceedings{
-  MMT,
-  title={Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification},
-  author={Yixiao Ge and Dapeng Chen and Hongsheng Li},
-  booktitle={International Conference on Learning Representations},
-  year={2020},
-  url={https://openreview.net/forum?id=rJlnOhVYPS}
+    MMT,
+    title={Mutual Mean-Teaching: Pseudo Label Refinery for Unsupervised Domain Adaptation on Person Re-identification},
+    author={Yixiao Ge and Dapeng Chen and Hongsheng Li},
+    booktitle={ICLR},
+    year={2020},
 }
 ```
diff --git a/examples/domain_adaptation/re_identification/ibn.sh b/examples/domain_adaptation/re_identification/ibn.sh
@@ -1,36 +1,36 @@
 #!/usr/bin/env bash
 # Market1501 -> Duke
 CUDA_VISIBLE_DEVICES=0 python baseline.py data data -s Market1501 -t DukeMTMC -a resnet50_ibn_a \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/Market2Duke
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/ibn/Market2Duke
 CUDA_VISIBLE_DEVICES=0 python baseline.py data data -s Market1501 -t DukeMTMC -a resnet50_ibn_b \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/Market2Duke
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/ibn/Market2Duke
 
 # Duke -> Market1501
 CUDA_VISIBLE_DEVICES=0 python baseline.py data data -s DukeMTMC -t Market1501 -a resnet50_ibn_a \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/Duke2Market
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/ibn/Duke2Market
 CUDA_VISIBLE_DEVICES=0 python baseline.py data data -s DukeMTMC -t Market1501 -a resnet50_ibn_b \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/Duke2Market
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/ibn/Duke2Market
 
 # Market1501 -> MSMT
 CUDA_VISIBLE_DEVICES=0 python baseline.py data data -s Market1501 -t MSMT17 -a resnet50_ibn_a \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/Market2MSMT
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/ibn/Market2MSMT
 CUDA_VISIBLE_DEVICES=0 python baseline.py data data -s Market1501 -t MSMT17 -a resnet50_ibn_b \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/Market2MSMT
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/ibn/Market2MSMT
 
 # MSMT -> Market1501
 CUDA_VISIBLE_DEVICES=0 python baseline.py data data -s MSMT17 -t Market1501 -a resnet50_ibn_a \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/MSMT2Market
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/ibn/MSMT2Market
 CUDA_VISIBLE_DEVICES=0 python baseline.py data data -s MSMT17 -t Market1501 -a resnet50_ibn_b \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/MSMT2Market
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/ibn/MSMT2Market
 
 # Duke -> MSMT
 CUDA_VISIBLE_DEVICES=0 python baseline.py data data -s DukeMTMC -t MSMT17 -a resnet50_ibn_a \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/Duke2MSMT
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/ibn/Duke2MSMT
 CUDA_VISIBLE_DEVICES=0 python baseline.py data data -s DukeMTMC -t MSMT17 -a resnet50_ibn_b \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/Duke2MSMT
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/ibn/Duke2MSMT
 
 # MSMT -> Duke
 CUDA_VISIBLE_DEVICES=0 python baseline.py data data -s MSMT17 -t DukeMTMC -a resnet50_ibn_a \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/MSMT2Duke
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/ibn/MSMT2Duke
 CUDA_VISIBLE_DEVICES=0 python baseline.py data data -s MSMT17 -t DukeMTMC -a resnet50_ibn_b \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/MSMT2Duke
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/ibn/MSMT2Duke
diff --git a/examples/domain_adaptation/re_identification/spgan.sh b/examples/domain_adaptation/re_identification/spgan.sh
@@ -4,44 +4,44 @@ CUDA_VISIBLE_DEVICES=0 python spgan.py data -s Market1501 -t DukeMTMC \
 --log logs/spgan/Market2Duke --translated-root data/spganM2D --seed 0
 # step2: train baseline on translated source dataset
 CUDA_VISIBLE_DEVICES=0 python baseline.py data/spganM2D data -s Market1501 -t DukeMTMC -a reid_resnet50 \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/Market2Duke
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/spgan/Market2Duke
 
 # Duke -> Market1501
 # step1: train SPGAN
 CUDA_VISIBLE_DEVICES=0 python spgan.py data -s DukeMTMC -t Market1501 \
 --log logs/spgan/Duke2Market --translated-root data/spganD2M --seed 0
 # step2: train baseline on translated source dataset
 CUDA_VISIBLE_DEVICES=0 python baseline.py data/spganD2M data -s DukeMTMC -t Market1501 -a reid_resnet50 \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/Duke2Market
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/spgan/Duke2Market
 
 # Market1501 -> MSMT17
 # step1: train SPGAN
 CUDA_VISIBLE_DEVICES=0 python spgan.py data -s Market1501 -t MSMT17 \
 --log logs/spgan/Market2MSMT --translated-root data/spganM2S --seed 0
 # step2: train baseline on translated source dataset
 CUDA_VISIBLE_DEVICES=0 python baseline.py data/spganM2S data -s Market1501 -t MSMT17 -a reid_resnet50 \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/Market2MSMT
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/spgan/Market2MSMT
 
 # MSMT -> Market1501
 # step1: train SPGAN
 CUDA_VISIBLE_DEVICES=0 python spgan.py data -s MSMT17 -t Market1501 \
 --log logs/spgan/MSMT2Market --translated-root data/spganS2M --seed 0
 # step2: train baseline on translated source dataset
 CUDA_VISIBLE_DEVICES=0 python baseline.py data/spganS2M data -s MSMT17 -t Market1501 -a reid_resnet50 \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/MSMT2Market
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/spgan/MSMT2Market
 
 # Duke -> MSMT
 # step1: train SPGAN
 CUDA_VISIBLE_DEVICES=0 python spgan.py data -s DukeMTMC -t MSMT17 \
 --log logs/spgan/Duke2MSMT --translated-root data/spganD2S --seed 0
 # step2: train baseline on translated source dataset
 CUDA_VISIBLE_DEVICES=0 python baseline.py data/spganD2S data -s DukeMTMC -t MSMT17 -a reid_resnet50 \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/Duke2MSMT
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/spgan/Duke2MSMT
 
 # MSMT -> Duke
 # step1: train SPGAN
 CUDA_VISIBLE_DEVICES=0 python spgan.py data -s MSMT17 -t DukeMTMC \
 --log logs/spgan/MSMT2Duke --translated-root data/spganS2D --seed 0
 # step2: train baseline on translated source dataset
 CUDA_VISIBLE_DEVICES=0 python baseline.py data/spganS2D data -s MSMT17 -t DukeMTMC -a reid_resnet50 \
---iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/baseline/MSMT2Duke
+--iters-per-epoch 800 --print-freq 80 --finetune --seed 0 --log logs/spgan/MSMT2Duke