KIMJUNGHOandClaude Opus 4.6
41d9e2c0e0
Remove controlnet_aux dependency to eliminate unwanted warnings (mediapipe, timm, segment_anything) that fire on every import. Each preprocessor is now implemented directly in preprocessors/ with only the needed dependencies. - canny: pure cv2.Canny() - depth: transformers DPTForDepthEstimation (Intel/dpt-hybrid-midas) - lineart: Generator model with HuggingFace weights (sk_model.pth) - openpose: body + hand detection extracted from controlnet_aux source - mlsd: MobileV2_MLSD_Large extracted from controlnet_aux source - Auto GPU detection (CUDA > MPS > CPU) for all neural network models Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
31 lines
1011 B
Python
31 lines
1011 B
Python
import cv2
|
|
import numpy as np
|
|
|
|
|
|
def HWC3(x):
|
|
"""Ensure image is in HWC 3-channel format."""
|
|
assert x.dtype == np.uint8
|
|
if x.ndim == 2:
|
|
return np.stack([x] * 3, axis=2)
|
|
if x.shape[2] == 1:
|
|
return np.concatenate([x] * 3, axis=2)
|
|
if x.shape[2] == 4:
|
|
color = x[:, :, :3].astype(np.float32)
|
|
alpha = x[:, :, 3:4].astype(np.float32) / 255.0
|
|
result = (color * alpha + 255.0 * (1.0 - alpha)).clip(0, 255).astype(np.uint8)
|
|
return result
|
|
return x
|
|
|
|
|
|
def resize_image(input_image, resolution):
|
|
"""Resize image maintaining aspect ratio, rounding to 64px boundaries."""
|
|
H, W = input_image.shape[:2]
|
|
k = float(resolution) / min(H, W)
|
|
H_new = int(np.round(H * k / 64.0)) * 64
|
|
W_new = int(np.round(W * k / 64.0)) * 64
|
|
if max(H_new, W_new) > max(H, W):
|
|
interp = cv2.INTER_LANCZOS4
|
|
else:
|
|
interp = cv2.INTER_AREA
|
|
return cv2.resize(input_image, (W_new, H_new), interpolation=interp)
|