MBCTD

A deep learning model for per-pixel, multi-label detection of building changes in bi-temporal satellite and aerial imagery. MBCTD classifies each pixel independently into three change categories, allowing overlapping labels (e.g., a replacement site marked as both demolished and new simultaneously).

---

Overview

Given a before image and an after image of the same geographic area, MBCTD produces per-pixel predictions for three classes:

Class	Color	Meaning
Unchanged	🔵	Building present in both images
Demolished	🔴	Building present in before, absent in after
New	🟢	Building absent in before, present in after
Replacement (demolished + new)	🟡	Both demolished and new labels active simultaneously

Because the model is multi-label, a single pixel can belong to more than one class. This makes it possible to represent complex urban transitions that single-label models cannot express.

Inference samples

LEVIR-CD+

FOTBCD

---

Architecture

MBCTD uses a Siamese ConvNeXt-Base encoder paired with a full-resolution U-Net decoder:

Before image ──┐
               ├─► Shared ConvNeXt-Base encoder ──► Change fusion at each scale
After image  ──┘                                           │
                                                           ▼
                                               Full-resolution U-Net decoder
                                               (PixelShuffle upsampling)
                                                           │
                                                           ▼
                                          3 independent sigmoid heads per pixel

Key design decisions:

• Shared encoder weights — the same ConvNeXt trunk processes both images, making the feature space comparable by construction.
• Change fusion — at each encoder scale, before/after features are combined as [before, after, before−after, |before−after|] and projected through 1×1→3×3 convolutions.
• High-resolution skip connections — in addition to encoder skips (1/32 → 1/4), raw input images are injected at 1/2 and 1/1 resolution to preserve fine-grained boundary information.
• PixelShuffle upsampling — learned upsampling at every decoder stage avoids checkerboard artifacts.
• Pre-trained backbone — ConvNeXt-Base initialised with DINOv3 LVD1689M weights.

---

Project Structure

MBCTD/
├── model.py               # Model definition (encoder, fusion, decoder)
├── config.py              # MBCTDConfig dataclass
├── inference.py           # load_model, predict_patch, visualisation helpers
├── demo.py                # Interactive Gradio web demo
└── environment.yml        # Conda environment spec

---

Installation

1. Clone the repository

git clone git@github.com:abdelpy/MBCTD
cd MBCTD

2. Create the Conda environment

conda env create -f environment.yml
conda activate mbctd

3. Install PyTorch

Follow the official instructions to install PyTorch matching your CUDA version.

4. Download model weights

Pre-trained weights are available on Google Drive.

Usage

Interactive demo

The fastest way to try the model is the Gradio web interface:

python demo.py path/to/model.pth

Open the URL printed in your terminal. The UI lets you:

• Upload a before and after image pair
• Adjust the confidence threshold (0.1 – 0.9, default 0.7)
• Choose an inference mode:
  • patch — tile the image into 256 px patches and stitch predictions back (handles large images)
  • full — run at the image's original resolution
• Inspect the overlay on the after image, the colour mask, and per-class pixel coverage statistics

---

Programmatic inference

from PIL import Image
import numpy as np
import torch
from inference import load_model, predict_patch

device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
model = load_model("model.pth", device)

# Load images as uint8 RGB numpy arrays
before = np.array(Image.open("before.png").convert("RGB"))
after  = np.array(Image.open("after.png").convert("RGB"))

result = predict_patch(before, after, model, threshold=0.7)

predict_patch returns a dictionary:

Key	Shape	dtype	Description
binary	(3, H, W)	uint8	Per-class binary masks (unchanged / demolished / new)
class_map	(H, W)	uint8	Collapsed single-label class ID (0 = background, 1–4 = see colour table above)
overlay	(H, W, 3)	uint8	Semi-transparent colour overlay drawn on the after image
mask_rgb	(H, W, 3)	uint8	Solid-colour mask visualisation

---

Training

MBCTD was trained exclusively on FOTBCD — a large-scale, multi-label building change dataset — using 256 × 256 px patches drawn from over 220,000 before/after aerial image pairs across France.

FOTBCD — the dataset behind the model

FOTBCD is the first dataset with multi-label building change annotations as vector polygons covering demolished, new, and unchanged structures simultaneously. At 220k+ georeferenced pairs spanning diverse urban environments, it is an order of magnitude larger than existing change detection benchmarks — and the richness of its labels is what makes a model like MBCTD possible.

The dataset is available for licensing.
Whether you are building an urban monitoring platform, a real-estate analytics product, or a geospatial AI pipeline, FOTBCD gives you the ground truth that generic benchmarks cannot provide. Get in touch to discuss licensing terms.

---

Results

Binary change detection metrics (demolished OR new → "changed") are reported for both benchmarks to enable comparison; LEVIR-CD+ provides only binary ground truth so no per-class breakdown is available for it. For FOTBCD, which supports multi-label annotations, per-class IoU is also reported.
Inference threshold selected by best F1 on each benchmark.

LEVIR-CD+ (full-resolution inference, threshold = 0.75)

Metric	Value
Precision	0.7694
Recall	0.8137
F1	0.7909
IoU change	0.6541
mIoU	0.8180
OA	0.9825

FOTBCD (full-resolution inference, threshold = 0.70)

Binary change detection

Metric	Value
Precision	0.8948
Recall	0.9201
F1	0.9073
IoU change	0.8303
mIoU	0.9094
OA	0.9891

Per-class IoU

Class	IoU
Unchanged	0.7774
Demolished	0.8166
New	0.8198

---

License

This project is licensed under CC BY-NC 4.0