https://github.com/masterskepticista/rooftop-instance-segmentation
VGG-16 Model to Segment Rooftops from Aerial Imagery
https://github.com/masterskepticista/rooftop-instance-segmentation
instance-segmentation rooftop-segmentation tensorflow vgg16 vgg16-python
Last synced: 3 months ago
JSON representation
VGG-16 Model to Segment Rooftops from Aerial Imagery
- Host: GitHub
- URL: https://github.com/masterskepticista/rooftop-instance-segmentation
- Owner: MasterSkepticista
- Created: 2019-07-11T17:19:38.000Z (almost 6 years ago)
- Default Branch: master
- Last Pushed: 2022-11-21T22:01:11.000Z (over 2 years ago)
- Last Synced: 2023-03-08T22:44:06.127Z (over 2 years ago)
- Topics: instance-segmentation, rooftop-segmentation, tensorflow, vgg16, vgg16-python
- Language: Python
- Size: 2.22 MB
- Stars: 20
- Watchers: 2
- Forks: 4
- Open Issues: 3
-
Metadata Files:
- Readme: README.md
Awesome Lists containing this project
README
# Rooftop Instance Segmentation using TensorFlow
### Aerial Imagery Dataset provided by National Topographic Office of New Zealand
Sample Outputs (downsized):
.png)
.png)
.png)
Link: https://www.airs-dataset.com/
Extract the dataset completely, store in this format:
```bash
# Training images
root/data/train/image
# Labels
root/data/train/label
# Test samples
root/data/test
```
Model Used: VGG-16, Instance Segmentation
The script was written on older version of TensorFlow (1.15.x and lower).
Some newer python versions do not include the listing of older TF.
You can downgrade to Python 3.6.x to use it, or use a python virtualenv to install specific python binary (recommended).
Install deps with the following command.
```python
pip install -r requirements.txt
```
Trained on Nvidia Quadro GP100 with 16GB VRAM. Batch Size:2 Input resolution 3584x3584.
Training Time: 12 hours