How to crop image correctly to minimize glb size after slice_mesh_plane a scene?

[yuezhilanyi]

I successfully split a glb file into two parts (including textures), but after the split, each part retained a complete texture file. I want to reduce the size of the textures to decrease the glb file size. I tried to crop the textures based on the UV extremes while stretching the UVs to fit within the 0-1 range, and then assigned the new textures and UVs to the split files. However, there's an issue with the experiment results: the texture range has shifted slightly. The image below shows the original side of the truck.

The image below shows the side of the truck after the split. You can see several black areas have been added to the upper part of the side, which are closely adjacent to the truck's side on the texture.

The image below shows the back of the original truck.

The image below shows the back of the truck after the split, with the bottom part of the taillights showing less than in the original file.

Does anyone know what went wrong in the process? Thanks very much in advance.

here is the original image and my code.

import io
import trimesh 

from trimesh.intersections import slice_mesh_plane
from PIL import Image

sc = trimesh.load('CesiumMilkTruck.glb')
mesh = sc.geometry['Cesium_Milk_Truck']

clipped = slice_mesh_plane(mesh, [-1, 0, 0], [ -1.40848477,  2.05412436, -0.00354755])

# # read material image 
material = mesh.visual.material.to_simple()
data, name = material.to_obj()
image_data = data[name + '.png']
image = Image.open(io.BytesIO(image_data))
image.size

uv = clipped.visual.uv
uv_maxs = uv.max(axis=0)
uv_mins = uv.min(axis=0)
height, width = image.height, image.width
uv.shape, uv.max(axis=0), uv.min(axis=0), uv.mean(axis=0)

# # crop material image
left = width * uv_mins[0]
right = width * uv_maxs[0]
upper = height * uv_mins[1]
lower = height * uv_maxs[1]
left, upper, right, lower

cropped_image = image.crop((left, upper, right, lower))
right-left, lower-upper, cropped_image.size

# # normalize uv
uv_norm = (uv - uv_mins) / (uv_maxs - uv_mins)
uv_norm.max(axis=0), uv_norm.min(axis=0), uv_norm.mean(axis=0), uv_norm[:3], uv[:3]

# # write material image to mesh 
color_visuals = trimesh.visual.TextureVisuals(uv=uv_norm, image=cropped_image)
clipped.visual = color_visuals
clipped.show()