如何在pyrender中将场景渲染为图像后获取对象的像素坐标？

Question

我正在尝试获取使用pyrender渲染的对象的像素坐标(x，y)。其目的是获得该对象的边界框坐标。我使用OffScreenRenderer来渲染场景。

r = pyrender.OffscreenRenderer(viewport_width=640,
                            viewport_height=480,
                            point_size=1.0)
color, depth = r.render(scene)

可用信息: camera_to_world姿态矩阵

首先，我尝试绘制对象的质心，如下所示：

x, y = trimesh_mesh.centroid[:2]
height, width = image.shape[:2]
x = int(x * width)
y = int(y * height)
plt.imshow(image)
plt.scatter(x, y)
plt.show()

我得到了以下信息：image plot

类似地，同一对象的边界框图是here

有人知道如何获得渲染对象的精确质心和长方体坐标吗？谢谢。

Answer 1

您正在寻找bpy_extras.object_utils.world_to_camera_view。

像这样使用它：

import bpy
import bpy_extras

scene = bpy.context.scene
obj = bpy.context.object   # object you want the coordinates of
vertices = obj.data.vertices   # you will get the coordinates of its vertices

for v in vertices:
    # local to global coordinates
    co = v.co @ obj.matrix_world
    # calculate 2d image coordinates
    co_2d = bpy_extras.object_utils.world_to_camera_view(scene, camera, co)
    render_scale = scene.render.resolution_percentage / 100
    render_size = (
        int(scene.render.resolution_x * render_scale),
        int(scene.render.resolution_y * render_scale),
    )

    # this is the result
    pixel_coords = (co_2d.x * render_size[0],
                    co_2d.y * render_size[1])

    print("Pixel Coords:", (
          round(pixel_coords[0]),
          round(pixel_coords[1]),
    ))

请参阅搅拌机堆栈交换上的this answer