ComfyUI AI Faces - Photo Verification Node

ComfyUI AI Faces - Photo Verification Node

A ComfyUI custom node for automated face verification, designed to check if a person is clearly visible and suitable for passport-style photos. This node performs comprehensive facial analysis to ensure photo quality meets identification document standards.

Features

This node performs the following verifications:

🔍 Face Detection & Count

• Detects faces in the image using InsightFace
• Ensures exactly one person is present in the photo
• Rejects photos with no faces or multiple people

📏 Face Size Validation

• Checks if face occupies appropriate portion of the image
• Face must be between 10% and 70% of the total image area
• Ensures proper distance from camera for ID photo standards

🎯 Head Pose Analysis

• Validates head orientation using 3D pose estimation
• Checks pitch (up/down tilt) - max 30° deviation
• Checks yaw (left/right turn) - max 30° deviation
• Checks roll (side tilt) - max 30° deviation
• Ensures face is looking straight at camera

🌟 Image Quality Assessment

• Analyzes face detection confidence score
• Requires minimum 75% confidence for acceptable quality
• Ensures proper lighting and image sharpness

👁️ Facial Feature Visibility

• Uses BiSeNet face parsing to segment facial features
• Validates presence and visibility of:
  • Left and right eyebrows (min 0.4% of face area)
  • Left and right eyes (min 0.3% of face area, unless wearing glasses)
  • Nose (min 2.0% of face area)
  • Upper and lower lips (min 0.4% of face area each)

🚫 Occlusion Detection

• Detects and rejects photos with:
  • Non-transparent sunglasses or dark eyewear
  • Hats, caps, or head coverings
• Allows transparent prescription glasses

🔬 Glasses Transparency Check

• Analyzes glasses pixel brightness to determine transparency
• Uses 50-pixel brightness threshold to distinguish clear vs. dark glasses
• Automatically adjusts feature visibility requirements for glasses wearers

Installation

Method 1: Git Clone (Recommended)

1. Navigate to your ComfyUI custom nodes directory:

cd /path/to/ComfyUI/custom_nodes/

2. Clone this repository:

git clone https://github.com/your-username/comfyui-ai-faces.git

3. Install dependencies:

cd comfyui-ai-faces
pip install -r requirements.txt

Method 2: Manual Installation

1. Download the repository as ZIP and extract to your ComfyUI custom nodes folder
2. Install the required packages:

pip install opencv-python torch numpy torchvision requests onnxruntime insightface supervision Pillow matplotlib

Method 3: ComfyUI Manager

If you have ComfyUI Manager installed:

1. Open ComfyUI Manager
2. Search for "AI Faces" or "Photo Verification"
3. Click Install

Usage

1. Restart ComfyUI after installation
2. The node will appear in the node menu under Fotobudka > Photo Verification
3. Connect an image input to the node
4. The node will:
   • ✅ Pass through the image if verification succeeds
   • ❌ Throw an error with specific feedback if verification fails

Example Workflow

Load Image → Photo Verification → Save Image

The node acts as a quality gate - images that pass verification continue through your workflow, while rejected images stop the workflow with descriptive error messages.

Error Messages

The node provides specific, actionable feedback:

• Face Detection Issues:

  • "No face detected. Please ensure your face is clearly visible."
  • "Only one person should be in the photo. Please ensure you are alone in the image."

• Distance Problems:

  • "Move closer to the camera so your face occupies more of the image."
  • "Move away from the camera so your face occupies less of the image."

• Pose Issues:

  • "Turn your head more towards the camera (too much side rotation)."
  • "Don't tilt your head too much up or down."
  • "Don't tilt your head to the side."
  • "Unable to determine head position."

• Quality Issues:

  • "Take a photo with better lighting and ensure image sharpness."

• Occlusion Problems:

  • "Please remove your glasses"
  • "Please remove your hat or cap"
  • "Please remove your glasses and hat or cap"

• Feature Visibility:

  • "Please show your [specific facial features] more clearly"

Technical Details

Models Used

• InsightFace Buffalo_L: Face detection and 3D pose estimation
• BiSeNet: 19-class face parsing for feature segmentation
• Execution Providers: CPU and GPU (CUDA) support

Face Parsing Classes

The node uses 19 semantic classes for face parsing:

• Background, face skin, eyebrows, eyes, nose, mouth, lips, hair, glasses, hat, etc.

Requirements

• Python 3.8+
• PyTorch
• OpenCV
• InsightFace
• ONNX Runtime
• Other dependencies in requirements.txt

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• InsightFace for face detection and analysis
• BiSeNet for face parsing
• ComfyUI community for the amazing framework

---

For questions, issues, or feature requests, please open an issue on GitHub.