VESNET - Vessel Level Detection

A deep learning-based solution for detecting liquid levels in transparent vessels using computer vision.

Overview

VESNET (Vessel Segmentation Network) is designed to accurately detect and measure liquid levels in transparent containers. It uses a combination of semantic segmentation and computer vision techniques to:

1. Detect vessels in images
2. Identify liquid content
3. Calculate fill levels with percentage accuracy

Model Architecture

The system uses two main neural network models:

• Semantic Segmentation Model: A fully convolutional network with Pyramid Scene Parsing (PSP) based on ResNet101 backbone
• YOLO Models: Used for vessel detection and localization

Pre-trained Models

The repository includes the following pre-trained models (stored using Git LFS):

• checkpoints/TrainedModelWeiht1m_steps_Semantic_TrainedWithLabPicsAndCOCO_AllSets.torch - Main segmentation model
• checkpoints/yolo11n.pt - YOLO model for vessel detection
• checkpoints/yolo11s.pt - YOLO model variant

Directory Structure

├── checkpoints/          # Pre-trained model weights
├── data/                 # Data directory
│   ├── input_data/      # Input images
│   └── output_data/     # Processing results
├── src/                 # Source code
│   ├── CategoryDictionary.py
│   ├── model.py         # Neural network architecture
│   ├── prepare_kaggle_data.py
│   └── visualize.py
├── main.py             # Main inference script
├── cam.py              # Camera interface
└── live.py            # Real-time processing

Setup

1. Clone the repository:

git clone https://github.com/yourusername/VESNET.git
cd VESNET

2. Install dependencies:

pip install -r requirements.txt

Usage

Running Inference

# Run on a single image
python main.py --image path/to/image.jpg

# Run on a folder of images
python main.py --folder path/to/folder

# Run on the full dataset
python main.py --dataset

Live Camera Mode

python live.py

Results

The system outputs:

• Segmentation masks for vessels and liquids
• Calculated fill levels (continuous percentage and nearest discrete level)
• Visualization overlays
• JSON results with detailed measurements

License

[Add your chosen license here]

Citation

If you use this work, please cite:

[Add citation information if applicable]

Acknowledgments

[Add any acknowledgments or credits]