[new] fix bugs and new features

Author: singer-yang

1_end2end_5lines.py

@@ -33,31 +33,34 @@ def config():
     with open('configs/end2end_5lines.yml') as f:
         args = yaml.load(f, Loader=yaml.FullLoader)
 
-    # ==> Device
-    num_gpus = torch.cuda.device_count()
-    args['num_gpus'] = num_gpus
-    device = torch.device(f"cuda" if torch.cuda.is_available() else "cpu")
-    args['device'] = device
-
     # ==> Result folder
     characters = string.ascii_letters + string.digits
     random_string = ''.join(random.choice(characters) for i in range(4))
     result_dir = f'./results/' + datetime.now().strftime("%m%d-%H%M%S") + '-End2End' + '-' + random_string
     args['result_dir'] = result_dir
     os.makedirs(result_dir, exist_ok=True)
     print(f'Result folder: {result_dir}')
-    
-    # ==> Logger
-    set_logger(result_dir)
-    logging.info(args)
 
     # ==> Random seed
     set_seed(args['train']['seed'])
 
+    # ==> Logger
+    set_logger(result_dir)
+    # Log to wandb 
     if not args['DEBUG']:
-        # wandb init
         pass
-        
+
+    # ==> Device
+    num_gpus = torch.cuda.device_count()
+    args['num_gpus'] = num_gpus
+    device = torch.device(f"cuda" if torch.cuda.is_available() else "cpu")
+    logging.info(f'Using {num_gpus} {torch.cuda.get_device_name(0)} GPU(s)')
+    args['device'] = device
+
+    # ==> Save config
+    with open(f'{result_dir}/config.yml', 'w') as f:
+        yaml.dump(args, f)
+
     return args
 
 

deeplens/optics.py

@@ -124,6 +124,9 @@ def post_computation(self):
         avg_pupilz, avg_pupilx = self.entrance_pupil()
         self.fnum = self.foclen / avg_pupilx / 2
 
+        if self.r_last < 8.0:
+            self.is_cellphone = True
+
 
     def find_aperture(self):
         """ Find aperture by surfaces previous and next materials.
@@ -820,7 +823,8 @@ def render_compute_image(self, img, depth, scale, ray, point_pixel=True, train=T
         irr_img += img[...,idx_i, idx_j+1] * w_i * (1-w_j)
         irr_img += img[...,idx_i+1, idx_j+1] * (1-w_i) * (1-w_j)
 
-        I = (torch.sum(irr_img * ray.ra, -3) + 1e-9) / (torch.sum(ray.ra, -3) + 1e-6)
+        I = (torch.sum(irr_img * ray.ra, -3) + 1e-9) / (torch.sum(ray.ra, -3) + 1e-6)   # w/ vignetting correction 
+        # I = (torch.sum(irr_img * ray.ra, -3) + 1e-9) / ray.ra.shape[-3]   # w/o vignetting correction
 
         # ====> Add sensor noise
         if noise > 0:
@@ -1469,7 +1473,7 @@ def set_target_fov_fnum(self, hfov, fnum, imgh=None):
 
         self.foclen = self.calc_efl()
         aper_r = self.foclen / fnum / 2
-        self.surfaces[self.aper_idx].r = aper_r
+        self.surfaces[self.aper_idx].r = float(aper_r)
 
 
     # ---------------------------
@@ -2216,11 +2220,10 @@ def loss_ray_angle(self, target=0.7, depth=DEPTH):
     def loss_reg(self):
         """ An empirical regularization loss for lens design.
         """
-        # For spherical lens design
-        loss_reg = 0.1 * self.loss_infocus() + self.loss_self_intersec(dist_bound=0.5, thickness_bound=0.5)
-        
-        # For cellphone lens design, use 0.01 * loss_reg
-        # loss_reg = 0.1 * self.loss_infocus() + self.loss_ray_angle() + (self.loss_self_intersec() + self.loss_last_surf()) #+ self.loss_surface()
+        if self.is_cellphone:
+            loss_reg = 0.1 * self.loss_infocus() + self.loss_ray_angle() + (self.loss_self_intersec() + self.loss_last_surf())
+        else:
+            loss_reg = 0.1 * self.loss_infocus() + self.loss_self_intersec(dist_bound=0.5, thickness_bound=0.5)
 
         return loss_reg
 

deeplens/surfaces.py

@@ -275,13 +275,22 @@ def surface_sample(self, N=1000):
         o2 = torch.stack((x2,y2,z2), 1).to(self.device)
         return o2
 
+    def surface(self, x, y):
+        """ Calculate z coordinate of the surface at (x, y) with offset.
+            
+            This function is used in lens setup plotting.
+        """
+        x = x if torch.is_tensor(x) else torch.tensor(x).to(self.device)
+        y = y if torch.is_tensor(y) else torch.tensor(y).to(self.device)
+        return self.sag(x, y)
+    
     def surface_with_offset(self, x, y):
         """ Calculate z coordinate of the surface at (x, y) with offset.
             
             This function is used in lens setup plotting.
         """
-        x = torch.tensor(x).to(self.device) if type(x) is float else x
-        y = torch.tensor(y).to(self.device) if type(y) is float else y
+        x = x if torch.is_tensor(x) else torch.tensor(x).to(self.device)
+        y = y if torch.is_tensor(y) else torch.tensor(y).to(self.device)
         return self.sag(x, y) + self.d
 
     def max_height(self):

setup.py

@@ -23,6 +23,7 @@
         'transformers',
         'lpips',
         'einops',
+        'timm',
     ],
     license='Creative Commons Attribution-NonCommercial 4.0 International License',
     classifiers=[