Update optimal hyper parameters for FixMatch (VisDA)

Author: thucbx99

examples/domain_adaptation/image_classification/README.md

@@ -154,7 +154,7 @@ for three times and report their average accuracy.
 | AFN         | 76.1   | 75.0 | 95.6  | 56.2  | 81.3 | 69.8 | 93.0  | 81.0  | 93.4  | 74.1   | 91.7  | 55.0   | 90.6  | 18.1  | 74.4 |
 | MDD         | /      | 82.0 | 88.3  | 62.8  | 85.2 | 69.9 | 91.9  | 95.1  | 94.4  | 81.2   | 93.8  | 89.8   | 84.1  | 47.9  | 79.8 |
 | MCC         | 78.8   | 83.6 | 95.3  | 85.8  | 77.1 | 68.0 | 93.9  | 92.9  | 84.5  | 79.5   | 93.6  | 93.7   | 85.3  | 53.8  | 80.4 |
-| FixMatch    | /      | 77.5 | 94.3  | 75.1  | 72.4 | 87.2 | 95.6  | 88.5  | 92.2  | 37.5   | 97.3  | 84.6   | 88.0  | 17.0  | 77.1 |
+| FixMatch    | /      | 79.5 | 96.5  | 76.6  | 72.6 | 84.6 | 96.3  | 92.6  | 90.5  | 81.8   | 91.9  | 74.6   | 87.3  | 8.6   | 78.4 |
 
 ### DomainNet accuracy on ResNet-101
 

examples/domain_adaptation/image_classification/fixmatch.sh

@@ -22,5 +22,5 @@ CUDA_VISIBLE_DEVICES=0 python fixmatch.py data/office-home -d OfficeHome -s Rw -
 CUDA_VISIBLE_DEVICES=0 python fixmatch.py data/office-home -d OfficeHome -s Rw -t Pr -a resnet50 --lr 0.003 --bottleneck-dim 1024 --epochs 20 --seed 0 --log logs/fixmatch/OfficeHome_Rw2Pr
 
 # ResNet101, VisDA-2017, Single Source
-CUDA_VISIBLE_DEVICES=0 python fixmatch.py data/visda-2017 -d VisDA2017 -s Synthetic -t Real -a resnet101 \
-    --lr 0.003 --bottleneck-dim 2048 --epochs 20 -ub 64 --seed 0 --per-class-eval --train-resizing cen.crop --log logs/fixmatch/VisDA2017
+CUDA_VISIBLE_DEVICES=0 python fixmatch.py data/visda-2017 -d VisDA2017 -s Synthetic -t Real -a resnet101 --train-resizing cen.crop \
+    --lr 0.003 --threshold 0.8 --bottleneck-dim 2048 --epochs 20 -ub 64 --seed 0 --per-class-eval --log logs/fixmatch/VisDA2017