Add files via upload
This commit is contained in:
viviiii
@@ -1,6 +1,6 @@
|
||||
MIT License
|
||||
|
||||
Copyright (c) 2026 viviiii
|
||||
Copyright (c) 2025 ComfyUI-DepthNormalizer
|
||||
|
||||
Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
of this software and associated documentation files (the "Software"), to deal
|
||||
|
||||
@@ -1,2 +1,95 @@
|
||||
# ComfyUI-DepthNormalizer
|
||||
ComfyUI custom node to normalize depth maps into stable 0–190 8-bit range.
|
||||
|
||||
A ComfyUI custom node for normalizing depth maps to a specific range (0–190 by default).
|
||||
|
||||
## 简介
|
||||
|
||||
ComfyUI-DepthNormalizer 提供了一个自定义节点，用于将深度图按 0–255 灰阶域进行归一化和拉伸，使输出稳定在 0–190（可配置）。
|
||||
|
||||
**核心功能：**
|
||||
- 在 8-bit 域（0-255）中手动设置最小/最大值进行重映射
|
||||
- 支持将输出最大值映射到目标值（默认 190）
|
||||
- 可选减去最小值，强制黑色为 0
|
||||
- 自动处理 ComfyUI IMAGE 格式（0-1 float）与 8-bit 域的转换
|
||||
|
||||
## 安装方式
|
||||
|
||||
### 方式一：通过 ComfyUI Manager 安装（推荐）
|
||||
|
||||
1. 打开 ComfyUI，点击 **Manager** 按钮
|
||||
2. 选择 **Install Custom Nodes**
|
||||
3. 搜索 "DepthNormalizer" 或输入仓库地址
|
||||
4. 点击安装并重启 ComfyUI
|
||||
|
||||
### 方式二：手动安装
|
||||
|
||||
```bash
|
||||
# 进入 ComfyUI 的 custom_nodes 目录
|
||||
cd /path/to/ComfyUI/custom_nodes
|
||||
|
||||
# 克隆仓库
|
||||
git clone https://github.com/your-username/ComfyUI-DepthNormalizer.git
|
||||
|
||||
# 重启 ComfyUI
|
||||
```
|
||||
|
||||
## 节点说明
|
||||
|
||||
### Depth Range → 0–190 (Manual)
|
||||
|
||||
将深度图按指定的 8-bit 范围进行归一化，输出到目标范围。
|
||||
|
||||
#### 输入参数
|
||||
|
||||
| 参数名 | 类型 | 默认值 | 范围 | 说明 |
|
||||
|--------|------|--------|------|------|
|
||||
| `image` | IMAGE | - | - | 输入深度图（ComfyUI IMAGE 格式，0-1 float） |
|
||||
| `min_8bit` | INT | 0 | 0-255 | 输入深度图在 8-bit 域的最小值 |
|
||||
| `max_8bit` | INT | 190 | 0-255 | 输入深度图在 8-bit 域的最大值 |
|
||||
| `target_max_8bit` | INT | 190 | 1-255 | 输出深度图在 8-bit 域的目标最大值 |
|
||||
| `subtract_min` | BOOLEAN | True | - | 是否先减去最小值，使黑色强制为 0 |
|
||||
| `clamp_to_target` | BOOLEAN | True | - | 是否将输出限制在目标范围内 |
|
||||
|
||||
#### 输出
|
||||
|
||||
| 输出名 | 类型 | 说明 |
|
||||
|--------|------|------|
|
||||
| `image` | IMAGE | 归一化后的深度图（ComfyUI IMAGE 格式，0-1 float） |
|
||||
|
||||
#### 处理逻辑
|
||||
|
||||
1. 将输入从 0-1 float 转换为 0-255 域：`x255 = image * 255.0`
|
||||
2. 根据 `subtract_min` 参数选择重映射方式：
|
||||
- 如果为 True：`y255 = (x255 - min) * (target / (max - min))`
|
||||
- 如果为 False：`y255 = x255 * (target / max)`
|
||||
3. 根据 `clamp_to_target` 进行裁剪：
|
||||
- 如果为 True：限制在 `[0, target]`
|
||||
- 如果为 False：限制在 `[0, 255]`
|
||||
4. 转换回 0-1 float：`output = y255 / 255.0`
|
||||
|
||||
## 使用示例
|
||||
|
||||
### 示例：将深度范围 1-117 拉伸到 0-190
|
||||
|
||||
假设你有一个深度图，其有效深度值在 8-bit 域的范围是 1 到 117，你希望将其拉伸到 0-190 范围：
|
||||
|
||||
| 参数 | 值 |
|
||||
|------|-----|
|
||||
| `min_8bit` | 1 |
|
||||
| `max_8bit` | 117 |
|
||||
| `target_max_8bit` | 190 |
|
||||
| `subtract_min` | ✅ True |
|
||||
| `clamp_to_target` | ✅ True |
|
||||
|
||||
**效果：**
|
||||
- 输入值 1 将映射到 0
|
||||
- 输入值 117 将映射到 190
|
||||
- 中间值线性插值
|
||||
|
||||
## 分类
|
||||
|
||||
在 ComfyUI 节点树中位于：**Depth/Normalize**
|
||||
|
||||
## 许可证
|
||||
|
||||
MIT License
|
||||
|
||||
+16
@@ -0,0 +1,16 @@
|
||||
"""
|
||||
ComfyUI-DepthNormalizer
|
||||
A ComfyUI custom node for normalizing depth maps to a specific range.
|
||||
"""
|
||||
|
||||
from .nodes.depth_range_manual import DepthRangeTo190Manual
|
||||
|
||||
NODE_CLASS_MAPPINGS = {
|
||||
"DepthRangeTo190Manual": DepthRangeTo190Manual,
|
||||
}
|
||||
|
||||
NODE_DISPLAY_NAME_MAPPINGS = {
|
||||
"DepthRangeTo190Manual": "Depth Range → 0–190 (Manual)",
|
||||
}
|
||||
|
||||
__all__ = ["NODE_CLASS_MAPPINGS", "NODE_DISPLAY_NAME_MAPPINGS"]
|
||||
@@ -0,0 +1,3 @@
|
||||
from .depth_range_manual import DepthRangeTo190Manual
|
||||
|
||||
__all__ = ["DepthRangeTo190Manual"]
|
||||
@@ -0,0 +1,53 @@
|
||||
import torch
|
||||
|
||||
|
||||
class DepthRangeTo190Manual:
|
||||
"""
|
||||
Manual min/max remap in 8-bit domain (0..255) for depth maps.
|
||||
Output will be scaled so that max maps to target_max_8bit (default 190).
|
||||
Optionally subtract min first to force black = 0.
|
||||
"""
|
||||
|
||||
@classmethod
|
||||
def INPUT_TYPES(cls):
|
||||
return {
|
||||
"required": {
|
||||
"image": ("IMAGE",),
|
||||
"min_8bit": ("INT", {"default": 0, "min": 0, "max": 255, "step": 1}),
|
||||
"max_8bit": ("INT", {"default": 190, "min": 0, "max": 255, "step": 1}),
|
||||
"target_max_8bit": ("INT", {"default": 190, "min": 1, "max": 255, "step": 1}),
|
||||
"subtract_min": ("BOOLEAN", {"default": True}),
|
||||
"clamp_to_target": ("BOOLEAN", {"default": True}),
|
||||
}
|
||||
}
|
||||
|
||||
RETURN_TYPES = ("IMAGE",)
|
||||
RETURN_NAMES = ("image",)
|
||||
FUNCTION = "run"
|
||||
CATEGORY = "Depth/Normalize"
|
||||
|
||||
def run(self, image, min_8bit, max_8bit, target_max_8bit, subtract_min, clamp_to_target):
|
||||
x = image # ComfyUI IMAGE is typically float32 in 0..1
|
||||
|
||||
# Convert 0..1 float image to 0..255 domain
|
||||
x255 = x * 255.0
|
||||
|
||||
minv = float(min_8bit)
|
||||
maxv = float(max_8bit)
|
||||
target = float(target_max_8bit)
|
||||
|
||||
if subtract_min:
|
||||
denom = max(maxv - minv, 1e-6)
|
||||
y255 = (x255 - minv) * (target / denom)
|
||||
else:
|
||||
denom = max(maxv, 1e-6)
|
||||
y255 = x255 * (target / denom)
|
||||
|
||||
if clamp_to_target:
|
||||
y255 = torch.clamp(y255, 0.0, target)
|
||||
else:
|
||||
y255 = torch.clamp(y255, 0.0, 255.0)
|
||||
|
||||
# Back to 0..1 float for ComfyUI
|
||||
y = y255 / 255.0
|
||||
return (y,)
|
||||
Reference in New Issue
Block a user