enable changing crop bbox

panels/nerf_object_panel.py

@@ -2,7 +2,16 @@
 import math
 
 from turbo_nerf.blender_utility.driver_utility import force_update_drivers
-from turbo_nerf.blender_utility.obj_type_utility import get_active_nerf_obj, get_all_training_cam_objs, get_closest_parent_of_type, get_nerf_obj_type, get_nerf_training_cams, is_nerf_obj_type, is_self_or_some_parent_of_type
+
+from turbo_nerf.blender_utility.obj_type_utility import (
+    get_active_nerf_obj,
+    get_all_training_cam_objs,
+    get_closest_parent_of_type,
+    get_nerf_training_cams,
+    is_nerf_obj_type,
+    is_self_or_some_parent_of_type
+)
+
 from turbo_nerf.constants import (
     CAMERA_FAR_ID,
     CAMERA_INDEX_ID,
@@ -13,10 +22,13 @@
     OBJ_TYPE_NERF,
     OBJ_TYPE_TRAIN_CAMERA
 )
+from turbo_nerf.utility.math import clamp
 from turbo_nerf.utility.nerf_manager import NeRFManager
 from turbo_nerf.utility.render_camera_utils import bl2nerf_cam_train
 
-# general camera prop getter
+# helper methods
+# kinda dirty to put these in the global scope for this file
+
 def get_props_for_cams(nerf_obj, prop_name, default_value):    
     cam_objs = get_nerf_training_cams(nerf_obj, bpy.context)
 
@@ -25,16 +37,13 @@ def get_props_for_cams(nerf_obj, prop_name, default_value):
 
     return [o[prop_name] for o in cam_objs]
 
-# kindof a hack for now, need to figure out if there is a cleaner way to do this
-# i also don't like that it's just a global method in this file
-def update_dataset_cams(nerf_obj):
-    nerf = NeRFManager.get_nerf_for_obj(nerf_obj)
+# updates the dataset cameras for a nerf object
+def set_props_for_cams(nerf_obj, nerf):
     cam_objs = get_all_training_cam_objs(nerf_obj)
 
     # todo: only update the cameras that have changed
     nerf.dataset.cameras = [bl2nerf_cam_train(cam_obj, relative_to=nerf_obj) for cam_obj in cam_objs]
     nerf.is_dataset_dirty = True
-
 
 # Custom property group
 class NeRFObjectProperties(bpy.types.PropertyGroup):
@@ -78,7 +87,7 @@ def set_near(self, value):
             o[CAMERA_NEAR_ID] = value
             force_update_drivers(o)
 
-        update_dataset_cams(nerf_obj)
+        set_props_for_cams(nerf_obj, NeRFManager.get_nerf_for_obj(nerf_obj))
 
     near: bpy.props.FloatProperty(
         name="Near Plane",
@@ -104,7 +113,7 @@ def set_far(self, value):
             o[CAMERA_FAR_ID] = value
             force_update_drivers(o)
 
-        update_dataset_cams(nerf_obj)
+        set_props_for_cams(nerf_obj, NeRFManager.get_nerf_for_obj(nerf_obj))
 
     far: bpy.props.FloatProperty(
         name="Far Plane",
@@ -125,12 +134,13 @@ def get_show_image_planes(prop_id):
 
     def set_show_image_planes(self, value):
         nerf_obj = get_active_nerf_obj(bpy.context)
+
         cams = get_nerf_training_cams(nerf_obj, bpy.context)
         for o in cams:
             o[CAMERA_SHOW_IMAGE_PLANES_ID] = value
             force_update_drivers(o)
 
-        update_dataset_cams(nerf_obj)
+        set_props_for_cams(nerf_obj, NeRFManager.get_nerf_for_obj(nerf_obj))
 
     show_image_planes: bpy.props.BoolProperty(
         name="Is Visible",
@@ -140,6 +150,84 @@ def set_show_image_planes(self, value):
         set=set_show_image_planes,
     )
 
+    # cropping (render bbox)
+
+    def get_crop(dim: str):
+        if dim not in ("x", "y", "z"):
+            raise ValueError(f"Invalid dimension {dim} for crop")
+
+        def crop_getter(prop_id):
+            nerf_obj = get_active_nerf_obj(bpy.context)
+            nerf = NeRFManager.get_nerf_for_obj(nerf_obj)
+            min_dim = getattr(nerf.render_bbox, f"min_{dim}")
+            max_dim = getattr(nerf.render_bbox, f"max_{dim}")
+            return (min_dim, max_dim)
+
+        return crop_getter
+
+    def set_crop(dim):
+        if dim not in ("x", "y", "z"):
+            raise ValueError(f"Invalid dimension {dim} for crop")
+
+        def crop_setter(self, value):
+            nerf_obj = get_active_nerf_obj(bpy.context)
+            nerf = NeRFManager.get_nerf_for_obj(nerf_obj)
+
+            tbbox_min = getattr(nerf.training_bbox, f"min_{dim}")
+            tbbox_max = getattr(nerf.training_bbox, f"max_{dim}")
+
+            rbbox_min = getattr(nerf.render_bbox, f"min_{dim}")
+            rbbox_max = getattr(nerf.render_bbox, f"max_{dim}")
+
+            new_min = clamp(value[0], tbbox_min, rbbox_max)
+            new_max = clamp(value[1], rbbox_min, tbbox_max)
+
+            new_bbox = nerf.render_bbox
+            setattr(new_bbox, f"min_{dim}", new_min)
+            setattr(new_bbox, f"max_{dim}", new_max)
+            nerf.render_bbox = new_bbox
+
+            force_update_drivers(nerf_obj)
+
+        return crop_setter
+
+    crop_x: bpy.props.FloatVectorProperty(
+        name="Crop X",
+        description="Crop X",
+        min=-128.0,
+        max=128.0,
+        size=2,
+        step=0.01,
+        precision=5,
+        get=get_crop("x"),
+        set=set_crop("x"),
+    )
+
+    crop_y: bpy.props.FloatVectorProperty(
+        name="Crop Y",
+        description="Crop Y",
+        min=-128.0,
+        max=128.0,
+        size=2,
+        step=0.01,
+        precision=5,
+        get=get_crop("y"),
+        set=set_crop("y"),
+    )
+
+    crop_z: bpy.props.FloatVectorProperty(
+        name="Crop Z",
+        description="Crop Z",
+        min=-128.0,
+        max=128.0,
+        size=2,
+        step=0.01,
+        precision=5,
+        get=get_crop("z"),
+        set=set_crop("z"),
+    )
+
+
 # Custom panel for the dropdown
 class NeRFObjectPanel(bpy.types.Panel):
     bl_label = "TurboNeRF Properties"
@@ -197,12 +285,19 @@ def draw_camera_section(self, context, ui_props):
         row = layout.row()
         row.prop(ui_props, "show_image_planes")
 
-        # row = layout.row()
-        # row.prop(ui_props, "use_for_training")
+        # crop section
+
+        row = layout.row()
+        row.label(text="Crop")
+
+        row = layout.row()
+        row.prop(ui_props, "crop_x")
 
-        # row = layout.row()
-        # row.prop(ui_props, "focal_length")
+        row = layout.row()
+        row.prop(ui_props, "crop_y")
 
+        row = layout.row()
+        row.prop(ui_props, "crop_z")
 
     # Register the custom property group and panel
     @classmethod

panels/nerf_panel_operators/load_nerf_images_operator.py

@@ -2,7 +2,6 @@
 
 from turbo_nerf.blender_utility.blender_ui_utility import switch_to_turbo_nerf_renderer
 from turbo_nerf.blender_utility.obj_type_utility import get_active_nerf_obj
-from turbo_nerf.constants import NERF_ITEM_IDENTIFIER_ID
 from turbo_nerf.utility.nerf_manager import NeRFManager
 
 class LoadNeRFImagesOperator(bpy.types.Operator):

renderer/nerf_render_engine.py

@@ -36,7 +36,6 @@ def __init__(self):
         self.cancel_current_render = False
         self.current_region3d: bpy.types.RegionView3D = None
         self.latest_camera = None
-        self.has_depsgraph_updates = True
 
         self.prev_view_dims = (0, 0)
 
@@ -263,15 +262,13 @@ def view_update(self, context, depsgraph: bpy.types.Depsgraph):
                 if nerf_obj_type is None:
                     continue
 
-                # Base NeRF object
+                # Update the NeRF representation's transform on the CUDA side
                 if nerf_obj_type == OBJ_TYPE_NERF:
                     nerf_id = obj[NERF_ITEM_IDENTIFIER_ID]
                     nerf = NeRFManager.get_nerf_by_id(nerf_id)
                     # why do we need to multiply by NERF_ADJUSTMENT_MATRIX? idk, but it works.
                     mat = np.array(obj.matrix_world @ NERF_ADJUSTMENT_MATRIX)
                     nerf.transform = tn.Transform4f(mat).from_nerf()
-
-                    self.has_depsgraph_updates = True
 
     # For viewport renders, this method is called whenever Blender redraws
     # the 3D viewport. The renderer is expected to quickly draw the render
@@ -309,12 +306,11 @@ def view_draw(self, context, depsgraph):
 
         has_new_camera = True if self.latest_camera is None else camera != self.latest_camera
         has_new_dims = dimensions != self.prev_view_dims
-        is_any_dataset_dirty =  np.any([nerf.is_dataset_dirty and nerf.can_render for nerf in all_nerfs])
+        is_any_nerf_dirty =  np.any([nerf.is_dirty() and nerf.can_render for nerf in all_nerfs])
 
-        user_initiated = has_new_camera or has_new_dims or is_any_dataset_dirty
+        user_initiated = has_new_camera or has_new_dims or is_any_nerf_dirty
 
-        if user_initiated or self.has_depsgraph_updates:
-            self.has_depsgraph_updates = False
+        if user_initiated:
             flags = tn.RenderFlags.Preview
             if not NeRFManager.is_training(): # and not context.screen.is_animation_playing:
                 flags = flags | tn.RenderFlags.Final

utility/math.py

@@ -31,3 +31,11 @@ def bl2nerf_pos(
     ) -> np.array:
     xyz_cycled = np.array([xyz[1], xyz[2], xyz[0]])
     return scale * xyz_cycled + origin
+
+def clamp(x, min, max):
+    if x < min:
+        return min
+    elif x > max:
+        return max
+    else:
+        return x

utility/nerf_manager.py

@@ -197,4 +197,3 @@ def getter(self):
             return cls.get_bridge_object_property(obj_name, prop_name, default)
 
         return getter
-