bash preprocess.sh
If you don't modify any hyperparameter,
python train.py
Or, store hyperparameters in Exps folder as json file. (Ex. hparam0.json {"gpu": "cuda:7"})
python train.py --hparam Exps/hparam0.json
All results will be stored in Results/Result{result_num}
Hyperparameters: "Results/Result{result_num}/hparam.json"
Training status: "Results/Result{result_num}/train_log.txt"
Images from each evaluation: "Results/Result{result_num}/Samples/"
You can plot graphs by entering following command.
python plot.py --result {result_num}
You can check 200 samples from test set and evaluate lpips, psnr, ssim.
(If you are still running the training code, I recommend you to generate new hyperparameter json file to avoid GPU memory shortage. Ex. Results/Result{result_num}/hparam_test.json {..., "gpu": "cuda:8", ...})
python reviewer.py --hparam Results/Result{result_num}/hparam.json --test 1
Testing status: "Results/Result{result_num}/test_log.txt"
Rendered images: "Results/Result{result_num}/Tested/"
You can generate a video.
(If you are still running the training code, I recommend you to generate new hyperparameter json file to avoid GPU memory shortage. Ex. Results/Result{result_num}/hparam_render.json {..., "gpu": "cuda:9", ...})
python reviewer.py --hparam Results/Result{result_num}/hparam.json --test 0
Rendering status: "Results/Result{result_num}/render_log.txt"
Rendered images: "Results/Result{result_num}/Rendered/"
Video: "Results/Result{result_num}/rendered.mp4"