Add a FID evaluator for stylegan with structural guidance

codes/models/eval/__init__.py

@@ -1,3 +1,4 @@
+from models.eval.sr_style import SrStyleTransferEvaluator
 from models.eval.style import StyleTransferEvaluator
 
 
@@ -5,5 +6,7 @@ def create_evaluator(model, opt_eval, env):
     type = opt_eval['type']
     if type == 'style_transfer':
         return StyleTransferEvaluator(model, opt_eval, env)
+    elif type == 'sr_stylegan':
+        return SrStyleTransferEvaluator(model, opt_eval, env)
     else:
         raise NotImplementedError()

codes/models/eval/sr_style.py

@@ -0,0 +1,54 @@
+import os
+
+import torch
+import os.path as osp
+import torchvision
+from torch.utils.data import BatchSampler
+
+import models.eval.evaluator as evaluator
+from pytorch_fid import fid_score
+
+
+# Evaluate that feeds a LR structure into the input, then calculates a FID score on the results added to
+# the interpolated LR structure.
+from data.stylegan2_dataset import Stylegan2Dataset
+
+
+class SrStyleTransferEvaluator(evaluator.Evaluator):
+    def __init__(self, model, opt_eval, env):
+        super().__init__(model, opt_eval, env)
+        self.batches_per_eval = opt_eval['batches_per_eval']
+        self.batch_sz = opt_eval['batch_size']
+        self.im_sz = opt_eval['image_size']
+        self.scale = opt_eval['scale']
+        self.fid_real_samples = opt_eval['real_fid_path']
+        self.gen_output_index = opt_eval['gen_index'] if 'gen_index' in opt_eval.keys() else 0
+        self.dataset = Stylegan2Dataset({'path': self.fid_real_samples,
+                                         'target_size': self.im_sz,
+                                         'aug_prob': 0,
+                                         'transparent': False})
+        self.sampler = BatchSampler(self.dataset, self.batch_sz, False)
+
+    def perform_eval(self):
+        fid_fake_path = osp.join(self.env['base_path'], "..", "fid_fake", str(self.env["step"]))
+        os.makedirs(fid_fake_path, exist_ok=True)
+        fid_real_path = osp.join(self.env['base_path'], "..", "fid_real", str(self.env["step"]))
+        os.makedirs(fid_real_path, exist_ok=True)
+        counter = 0
+        for batch in self.sampler:
+            noise = torch.FloatTensor(self.batch_sz, 3, self.im_sz, self.im_sz).uniform_(0., 1.).to(self.env['device'])
+            batch_hq = [e['GT'] for e in batch]
+            batch_hq = torch.stack(batch_hq, dim=0).to(self.env['device'])
+            resized_batch = torch.nn.functional.interpolate(batch_hq, scale_factor=1/self.scale, mode="area")
+            gen = self.model(noise, resized_batch)
+            if not isinstance(gen, list) and not isinstance(gen, tuple):
+                gen = [gen]
+            gen = gen[self.gen_output_index]
+            out = gen + torch.nn.functional.interpolate(resized_batch, scale_factor=self.scale, mode='bilinear')
+            for b in range(self.batch_sz):
+                torchvision.utils.save_image(out[b], osp.join(fid_fake_path, "%i_.png" % (counter)))
+                torchvision.utils.save_image(batch_hq[b], osp.join(fid_real_path, "%i_.png" % (counter)))
+                counter += 1
+
+        return {"fid": fid_score.calculate_fid_given_paths([fid_real_path, fid_fake_path], self.batch_sz, True,
+                                                           2048)}

codes/train.py

@@ -291,7 +291,7 @@ class Trainer:
 
 if __name__ == '__main__':
     parser = argparse.ArgumentParser()
-    parser.add_argument('-opt', type=str, help='Path to option YAML file.', default='../options/train_stylegan2_faster.yml')
+    parser.add_argument('-opt', type=str, help='Path to option YAML file.', default='../options/train_stylegan2_celebA_separated_disc.yml')
     parser.add_argument('--launcher', choices=['none', 'pytorch'], default='none', help='job launcher')
     parser.add_argument('--local_rank', type=int, default=0)
     args = parser.parse_args()