- Team Name: VLL
- Team Members:
- Chen, Yu-Hsuan
- Ji, Yan-Yang
- Lai, Yung-Hsuan
- Date: 2022/06/17
Follow the steps below to set up environment
git clone git@github.com:Franklin905/110-2-ComputerVision-FinalProject.gitcd 110-2-ComputerVision-FinalProject- Place data.zip under 110-2-ComputerVision-FinalProject/ and run
unzip data.zip cd data- Place aflw_test.zip under data/, which is generated from the above step, and run
unzip aflw_test.zip cd ../- Set up an environment of python 3.8 (strongly recommended) and run
pip3 install -r requirements.txt- 7.1 If python version isn't 3.8, you might encounter error when loading .pkl file with pickle package. We guess TAs probably compressed the data in python 3.8 environment, so we recommend installing python 3.8. We encounter the error when running in python 3.6 environment.
- 7.2 Possible solution: run the command: pip3 install pickle5. Modify the second line, "import pickle", in data.py to "import pickle5 as pickle"
python3 DML_train.py --data_dir ./data --train_batch 64 --n_epoch 150 --lr_min 0.0025 --save_model_name rConvNext.pth12255average_model_best.pth is our best model trained by DML and weight avg. Predicted facial landmarks of testing data will be written in solution.txt.
python3 test.py --data_dir ./data --train_batch 128 --save_model_name 12255average_model_best.pth50 visualized pictures for training data, validation data, testing data each will be saved in vis_folder.
python3 visualization.py --data_dir ./data --save_dir vis_folder --save_model_name 12255average_model_best.pth --num_pic 50- OS: Ubuntu 20.04.2 LTS
- CPU: Intel(R) Core(TM) i7-9700K CPU @ 3.60GHz
- GPU: Nvidia TITAN RTX 24GB
If you have any problems reproducing our work, please contact us by e-mail.
- Chen, Yu-Hsuan : r10942088@ntu.edu.tw
- Ji, Yan-Yang : r10942090@ntu.edu.tw
- Lai, Yung-Hsuan: r10942097@ntu.edu.tw