The following packages are used in this homework
numpy==1.19.3
torch==1.9.0+cu111
torchvision==0.10.0+cu111
.
├──dataset
├──Testing
├──XXXX.jpg
├──Training
├──XXXX.jpg
├──classes.txt # contain 200 bird species and it's number (e.g. 001.Black_footed_Albatross)
├──training_labels.txt # filename and label mapping (e.g. 4283.jpg 115.Brewer_Sparrow)
├──testing.txt # test filename (e.g. 4282.jpg)
├──fold{i}.txt # we split training data into 5 fold (i=1,2,3,4,5)
├──functions # functions inside
├──saved_model # trained model weights
├──split_folds.py # split training dataset into 5 folds (fold{i}.txt)
├──train.py
├──inference.py
└──README.md
If you want to change the hyper-parameter, modify a few lines (line 18 to 29) of content in train.py
Note: FoldID = i indicates that the training dataset in fold{i}.txt will be the validation dataset
To run this code, use the following command
python train.py
After running train.py, the model weight with the highest accuracy in that FoldID will be saved in saved_model
The format of the file name is fold=X_ep=YY_acc=0.ZZZZ.pth
inference.py will catch the model weight in saved_model automatically. To reproduce the testing result, run
python inference.py
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent