26 lines
843 B
Markdown
26 lines
843 B
Markdown
# Resolution estimation (ML)
|
|
|
|
Resolution estimation can be done with a recent deep learning model by D. Mendez et al.
|
|
(see [Acta Cryst D, 80, 26-43](https://doi.org/10.1107/S2059798323010586)), adapted to Jungfraujoch.
|
|
Model used in the original paper is located in the [resonet/](resonet) directory, after converting to TorchScript format.
|
|
|
|
To simplify deployment the model runs via Python HTTP server.
|
|
|
|
## Installation
|
|
```
|
|
cd resonet
|
|
python -m venv .venv
|
|
source .venv/bin/activate
|
|
pip install -r requirements.txt
|
|
```
|
|
|
|
## Execution
|
|
```
|
|
cd resonet
|
|
source .venv/bin/activate
|
|
python jfjoch_inference_server.py --port=<UDP port, default `8000`> --dev=<device, e.g. `cpu` or `cuda:0` for 1st GPU, default is `cuda`>
|
|
```
|
|
|
|
## Known issues
|
|
The feature is experimental, and for the time being performance is low, below 100 Hz with 4 x Nvidia L4 GPUs.
|