wangbo/ComfyUI
1
0
Fork 0
mirror of https://github.com/comfyanonymous/ComfyUI.git synced 2026-05-15 03:27:24 +08:00
Code Issues Actions 25 Packages Projects Releases Wiki Activity

Modify upscaling to be closer to original CORE-135

Browse Source
This commit is contained in:
Talmaj Marinc 2026-05-11 21:19:28 +02:00
parent 2a20db3eb3
commit b296c6a1aa
1 changed files with 14 additions and 8 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
comfy/ldm/depth_anything_3/preprocess.py
View File

@ -5,12 +5,13 @@
# src/depth_anything_3/utils/alignment.py (sky-aware depth clip) # src/depth_anything_3/utils/alignment.py (sky-aware depth clip)
# src/depth_anything_3/model/da3.py::_process_mono_sky_estimation # src/depth_anything_3/model/da3.py::_process_mono_sky_estimation
# #
# We deliberately do NOT replicate the upstream cv2-based resize path. ComfyUI # Resize: ``comfy.utils.common_upscale`` with ``upscale_method="lanczos"``.
# already provides ``comfy.utils.common_upscale`` for high-quality bilinear # Upstream uses cv2 INTER_CUBIC (upscale) / INTER_AREA (downscale); a sweep
# resampling; using it keeps everything on-device and consistent with other # across {bilinear, bicubic, area, lanczos, bislerp} on a 768->504 test image
# ComfyUI preprocessors. The bilinear approximation is sufficient for the # showed lanczos has the lowest max-abs-diff vs the upstream cv2 output
# downstream depth-estimation task (verified visually against the upstream # (~0.13 vs 0.21-0.71 for the others), so we use it in both directions for
# bicubic path -- depth maps are virtually identical). # simplicity. This keeps the path stateless, on-device, and free of any
# OpenCV dependency.
from __future__ import annotations from __future__ import annotations
@ -83,8 +84,13 @@ def preprocess_image(
# (B, H, W, 3) -> (B, 3, H, W) # (B, H, W, 3) -> (B, 3, H, W)
x = image.movedim(-1, 1).contiguous() x = image.movedim(-1, 1).contiguous()
if (target_h, target_w) != (H, W): if (target_h, target_w) != (H, W):
# common_upscale takes a (B, C, H, W) tensor. # Upstream uses cv2 INTER_CUBIC (upscale) / INTER_AREA (downscale).
x = comfy.utils.common_upscale(x, target_w, target_h, "bilinear", "disabled") # Lanczos in ``common_upscale`` is anti-aliased and produces the
# closest pixel-wise match in a sweep across {bilinear, bicubic,
# area, lanczos, bislerp}. Used in both directions for simplicity.
x = comfy.utils.common_upscale(
x.float(), target_w, target_h, "lanczos", "disabled",
)
x = x.clamp(0.0, 1.0) x = x.clamp(0.0, 1.0)
mean = _IMAGENET_MEAN.to(device=x.device, dtype=x.dtype).view(1, 3, 1, 1) mean = _IMAGENET_MEAN.to(device=x.device, dtype=x.dtype).view(1, 3, 1, 1)