gezhaoDL

请问下数据集文件夹构成

3

配置文件train_gfpgan_v1中： dataroot_gt: datasets/ffhq/ffhq_512 文件夹下存放512*512的png图片，FFHQ下载的1024*1024 resize而成，是这样吗？ 另外这两个文件夹里应该放什么样的数据？ dataroot_lq: datasets/faces/validation/input dataroot_gt: datasets/faces/validation/reference

how to train a model like stylegan2-ffhq-config-f.pd params？

1

I trained a model with [images256x256.tar](https://paddlegan.bj.bcebos.com/datasets/images256x256.tar).The ckpt.keys() are `['gen', 'disc', 'gen_ema']`， when stylegan2-ffhq-config-f.pd params.keys() are ` ['age', 'eyes_open', 'eye_distance', 'eye_eyebrow_distance', 'eye_ratio', 'gender', 'lip_ratio', 'mouth_open', 'mouth_ratio', 'nose_mouth_distance', 'nose_ratio', 'nose_tip', 'pitch', 'roll',...

How to modify the code to support multi category training

1

I successfully trained single category key point detection. Then i want to train the net with multi category. Where and How to modify the code in plugins? I didn't find...