User Guide

Overview

DeepSeis is a desktop application for passive-source / ambient noise seismic data processing. It provides an integrated GUI workflow from data import, slice preparation, QC, DEB directional energy balancing, cross-correlation retrieval, to NMO stack profile viewing.

Installation

Install dependencies in the project root directory:

pip install -r deepseis_app/requirements.txt

Launch the application:

python run.py

The main window title will be DeepSeis — Ambient Noise Cross-Correlation.

Data Loading

Supported Formats

• SEG-Y: .sgy / .segy files with IBM/IEEE float support
• MATLAB: .mat files
• SAC Directory: Date/station organized SAC file trees

SAC Directory Structure

DeepSeis supports both nested and flat directory structures:

# Nested (by date)
root/
├── 2024-08-01/
│   ├── 1001.GP2.2024-08-01.sac
│   └── 1002.GP2.2024-08-01.sac
├── 2024-08-02/
│   └── ...

# Flat
root/
├── 1001.GP2.2024-08-01.sac
├── 1001.GP2.2024-08-02.sac
└── ...

Loading Steps

1. Click Load SAC dir… or Load file…
2. Configure station range, date range, NT, filtering, and decimation
3. Click Load to generate cache and prepare slices

Processing Workflow

Recommended processing sequence:

1. Create or open a project
2. Load data via file import or SAC directory
3. Check Raw Data preview and parameters (NT / Panels / Receivers / dt)
4. Click Prepare Data to build slices/cache
5. Apply QC (manual review or AI labeling)
6. Optionally apply Slice Filter and DEB Filter
7. Run Cross-Correlation retrieval
8. Review virtual shots in Result tab
9. Optionally run NMO Stack and view in Profile tab
10. Export results as SEG-Y or images

Quality Control

Manual Review

Use Review / Label Slices… to visually inspect and label all prepared slices. Assign quality labels (C1, C2, C3) based on signal characteristics.

AI-Assisted Labeling

For AI batch labeling:

1. Prepare data first
2. Configure AI Server URL and API Key
3. Set classification thresholds (C1 / C2 / C3)
4. Click AI Label Batch…
5. Review and refine labels manually if needed

Slice Filter

Use labels to filter panels for cross-correlation:

• Select labels in Use labels (e.g., 1,2)
• Configure label range (e.g., 1:1-100,2:1-80)
• Click Apply Slice Filter then Use Slice Filter

Cross-Correlation

Before running cross-correlation, ensure:

• Data is prepared
• Slice Filter / DEB Filter status is not stale
• NT, Panels, Receivers, and dt match your data
• Thread count is set appropriately

Output

Default output shape: (4001, N_virtual × N_receivers)

For 51 receivers: 26 virtual shots × 51 receivers = 1326 output traces.

NMO Stack

Required Inputs

• Shot dir: Virtual shot or cross-correlation output directory
• Coords CSV: Receiver geometry with shot/receiver coordinates

Optional Inputs

• Statics file: Statics correction
• Velocity file: External velocity model

Key Parameters

Export

SEG-Y Export

Click Export SEG-Y… to save processing results in SEG-Y format for further interpretation in standard seismic software.

Image Export

Click Export Image… to save the current display view as PNG or PDF for publication and presentations.

Troubleshooting