Artificial Intelligence for digital REStoration of Cultural Heritage (AIRES) API

This is the GitHub repository of the code employed in Section 5 of the Article

A cloud-native application for digital restoration of Cultural Heritage using nuclear imaging: the AIRES-CH project App, Alessandro Bombini, Fernando Garcìa Avello-Bofìas, Chiara Ruberto, Francesco Taccetti

submitted to MDPI/computers as extended version of the conference paper:

Bombini, A., Anderlini, L., dell’Agnello, L., Giacomini, F., Ruberto, C., Taccetti, F. (2022). Hyperparameter Optimisation of Artificial Intelligence for Digital REStoration of Cultural Heritages (AIRES-CH) Models. In: Gervasi, O., Murgante, B., Misra, S., Rocha, A.M.A.C., Garau, C. (eds) Computational Science and Its Applications – ICCSA 2022 Workshops. ICCSA 2022. Lecture Notes in Computer Science, vol 13377. Springer, Cham. https://doi.org/10.1007/978-3-031-10536-4_7

winner of the best paper award at the Workshop on Advancements in Applied Machine-learning and Data Analytics (AAMDA) workshop at the the 22nd International Conference on Computational Science and Its Applications (ICCSA 2022).

This repository build the RESTful API for serving the trained AIRES-CH DNN models.

The AIRES-CH DNNs are accessible for inference via the THESPIAN-XRF web app; the DNNs are furnished via a RESTful API offering three routes:

1. /1D/, to perform the recolouring inference using the 1D model described in sec. 3.3.2;
2. /2D/, to perform the recolouring inference using the 2D model described in sec. 3.2;
3. /pixel/, to perform single-pixel inference; this branch was developed for the goal of offering real-time recolouring during measurements.

Getting started Veloci Raptor 03/14/15 As easily understandable, one of the main features we expect from our RESTful API (and DNN models) is a fast reply time (a short inference time). In order to optimise this aspect, we developed three APIs, with three different frameworks: FastAPI, Flask and NodeJS.

For more details, we refer to the aforementioned papers.

---

Extra: Deployement infos

Env variables

NGINX

NGINX_HOST_PORT=8443 NGINX_CERT=./NGINX/cert NGINX_ROOT=./NGINX

Flask4NGINX vars

FLASK_BASE_URL=/flask_aires NGINX_PROXY_PASS_FLASK=http://ip

CROW4NGINX vars

FLASK

In App

FLASK_PORT=5999 FLASK_HOST=0.0.0.0

NAME_OF_IMAGE_IN_FILE='img'

In Dockers

FLASK_ROOT_DIR=./Flask WORKDIR_PATH=/flask_app

GUNICORN_WORKERS_PER_CORE=1 GUNICORN_WORKER_CLASS=gthread GUNICORN_THREADS=4

CROW

AIRES MODELS

PATH_TO_AIRES_MODELS=./AIRES_MODELS/ NAME_AIRES_MODEL_1D=model_1D_multi_input.h5 NAME_AIRES_MODEL_2D=model_2D.h5