https://github.com/facebookresearch/pytorch_GAN_zoo
运行中出现的错误:
1、应该先启动visdom,然后再运行命令
python -m visdom.server
2、change the code in visualization .py 62 lines :
# checkPointDir = os.path.join(kwargs["dir"], name)
checkPointDir = os.path.join(kwargs["dir"])
3、the command of eval is:
python eval.py visualization -n pubchem -m PGAN --save_dataset data/output --size_dataset 10
the mean of -m and -n is:
parser.add_argument('-m', '--module', help="Module to evaluate, available modules: PGAN, PPGAN, DCGAN", type=str, dest="module") parser.add_argument('-n', '--name', help="Model's name", type=str, dest="name")
notice: 这里一定要注意,-n 是模型的名称“pubchem”而不是它的全称"pubchem_s5_i62000.pt"
得到输出:
4、change the code in inspiration_generation.py 338 lines :
# checkPointDir = os.path.join(kwargs["dir"], name) checkPointDir = os.path.join(kwargs["dir"])
python eval.py inspirational_generation -n pubchem -m PGAN --input_image data/input/1.jpg -f data/output/feature_extractor
https://towardsdatascience.com/molecule-synthesis-using-ai-10e0e1f89568