Merged

Author: half-potato

gaussian_renderer/ever.py

@@ -27,6 +27,7 @@
 import numpy as np
 from icecream import ic
 from scene.dataset_readers import ProjectionType
+from utils import camera_utils_zipnerf
 
 def get_ray_directions(H, W, focal, center=None, random=True):
     """

scene/dataset_readers.py

@@ -24,7 +24,6 @@
 from scene.gaussian_model import BasicPointCloud
 import enum
 import json
-import cv2
 
 class ProjectionType(enum.Enum):
   """Camera projection type (perspective pinhole, fisheye, or 360 pano)."""
@@ -158,8 +157,7 @@ def readColmapCameras(cam_extrinsics, cam_intrinsics, images_folder, metadata_pa
             exposure = 1
             iso = 100
             aperature = 1
-        image = cv2.imread(image_path)
-        image = cv2.cvtColor(image, cv2.COLOR_BGR2RGB)
+        image = None
 
         cam_info = CameraInfo(uid=uid, R=R, T=T, FovY=FovY, FovX=FovX, image=image,
                               image_path=image_path, image_name=image_name, width=width,

utils/camera_utils.py

@@ -13,12 +13,13 @@
 import numpy as np
 from utils.general_utils import PILtoTorch
 from utils.graphics_utils import fov2focal
-from icecream import ic
+from PIL import Image
 
 WARNED = False
 
 def loadCam(args, id, cam_info, resolution_scale):
-    orig_w, orig_h = cam_info.image.shape[1], cam_info.image.shape[0]
+    image = Image.open(cam_info.image_path)
+    orig_w, orig_h = image.size
 
     if args.resolution in [1, 2, 4, 8]:
         resolution = round(orig_w/(resolution_scale * args.resolution)), round(orig_h/(resolution_scale * args.resolution))
@@ -39,7 +40,7 @@ def loadCam(args, id, cam_info, resolution_scale):
         scale = float(global_down) * float(resolution_scale)
         resolution = (int(orig_w / scale), int(orig_h / scale))
 
-    resized_image_rgb = PILtoTorch(cam_info.image, resolution)
+    resized_image_rgb = PILtoTorch(image, resolution)
 
     gt_image = resized_image_rgb[:3, ...]
     loaded_mask = None

utils/general_utils.py

@@ -20,8 +20,8 @@ def inverse_sigmoid(x):
     return torch.log(x/(1-x))
 
 def PILtoTorch(pil_image, resolution):
-    resized_image_PIL = cv2.resize(pil_image, resolution)
-    # resized_image_PIL = pil_image.resize(resolution)
+    # resized_image_PIL = cv2.resize(pil_image, resolution)
+    resized_image_PIL = pil_image.resize(resolution)
     resized_image = torch.from_numpy(np.array(resized_image_PIL)) / 255.0
     if len(resized_image.shape) == 3:
         return resized_image.permute(2, 0, 1)