https://github.com/dmitryduev/braai
braai [Bogus/Real Adversarial AI]: Real-bogus astrophysical event classification for the Zwicky Transient Facility (ZTF) using deep learning
https://github.com/dmitryduev/braai
astronomy deep-learning ztf
Last synced: 16 days ago
JSON representation
braai [Bogus/Real Adversarial AI]: Real-bogus astrophysical event classification for the Zwicky Transient Facility (ZTF) using deep learning
- Host: GitHub
- URL: https://github.com/dmitryduev/braai
- Owner: dmitryduev
- License: mit
- Created: 2019-06-15T00:58:59.000Z (almost 6 years ago)
- Default Branch: master
- Last Pushed: 2023-10-30T22:24:25.000Z (over 1 year ago)
- Last Synced: 2025-05-06T19:13:14.562Z (16 days ago)
- Topics: astronomy, deep-learning, ztf
- Language: Jupyter Notebook
- Homepage:
- Size: 18.8 MB
- Stars: 19
- Watchers: 2
- Forks: 2
- Open Issues: 4
-
Metadata Files:
- Readme: readme.md
- License: LICENSE
Awesome Lists containing this project
README
# `braai` \[Bogus/Real Adversarial AI\]
## Real-bogus classification for the Zwicky Transient Facility using deep learning
Efficient automated detection of flux-transient, reoccurring flux-variable, and moving objects
is increasingly important for large-scale astronomical surveys. `braai` is a convolutional-neural-network,
deep-learning real/bogus classifier designed to separate genuine astrophysical events and objects
from false positive, or bogus, detections in the data of the [Zwicky Transient Facilty (ZTF)](https://ztf.caltech.edu),
a new robotic time-domain survey currently in operation at the Palomar Observatory in California, USA.
`braai` demonstrates a state-of-the-art performance as quantified by
its low false negative and false positive rates.
For details, please see [Duev et al. 2019, MNRAS, 489 (3), 3582-3590](https://academic.oup.com/mnras/article/489/3/3582/5554758).
[arXiv:1907.11259](https://arxiv.org/pdf/1907.11259.pdf)
### `braai` architecture
### Dataset
todo: plots
### Classifier performance
### Use `braai`
See [this jupyter notebook](https://github.com/dmitryduev/braai/blob/master/nb/braai_run.ipynb)
#### Edge TPU
### Transfer learning with `braai`
#### Jupyter/Colab
See [this jupyter notebook](https://github.com/dmitryduev/braai/blob/master/nb/braai_tl.ipynb), or
[](https://colab.research.google.com/github/dmitryduev/braai/blob/master/nb/braai_tl.ipynb)
### Train your own `braai`
#### Jupyter/Colab
See [this jupyter notebook](https://github.com/dmitryduev/braai/blob/master/nb/braai_train.ipynb), or
[](https://colab.research.google.com/github/dmitryduev/braai/blob/master/nb/braai_train.ipynb)
#### Docker
Build and launch the app container:
```bash
# without GPU support:
docker build --rm -t braai:cpu -f Dockerfile .
# with GPU support (requires nvidia-docker):
docker build --rm -t braai:gpu -f gpu.Dockerfile .
# run:
# without GPU support:
docker run -it --rm --name braai -v /path/to/store/data:/data braai:cpu
# with GPU support (requires nvidia-docker) exposing the first GPU:
docker run --runtime=nvidia -e NVIDIA_VISIBLE_DEVICES=0 -it --rm --name braai -v /path/to/store/data:/data braai:gpu
```
---
Train `braai`:
```bash
python /app/braai.py --t_stamp 20190614_003916 --model VGG6 --epochs 200 --patience 50 --batch_size 64 --verbose
```