gradio
/
CITATION.cff
45 строк · 1.9 Кб
1cff-version: 1.2.0
2message: Please cite this project using these metadata.
3title: "Gradio: Hassle-free sharing and testing of ML models in the wild"
4abstract: >-
5Accessibility is a major challenge of machine learning (ML).
6Typical ML models are built by specialists and require
7specialized hardware/software as well as ML experience to
8validate. This makes it challenging for non-technical
9collaborators and endpoint users (e.g. physicians) to easily
10provide feedback on model development and to gain trust in
11ML. The accessibility challenge also makes collaboration
12more difficult and limits the ML researcher's exposure to
13realistic data and scenarios that occur in the wild. To
14improve accessibility and facilitate collaboration, we
15developed an open-source Python package, Gradio, which
16allows researchers to rapidly generate a visual interface
17for their ML models. Gradio makes accessing any ML model as
18easy as sharing a URL. Our development of Gradio is informed
19by interviews with a number of machine learning researchers
20who participate in interdisciplinary collaborations. Their
21feedback identified that Gradio should support a variety of
22interfaces and frameworks, allow for easy sharing of the
23interface, allow for input manipulation and interactive
24inference by the domain expert, as well as allow embedding
25the interface in iPython notebooks. We developed these
26features and carried out a case study to understand Gradio's
27usefulness and usability in the setting of a machine
28learning collaboration between a researcher and a
29cardiologist.
30authors:
31- family-names: Abid
32given-names: Abubakar
33- family-names: Abdalla
34given-names: Ali
35- family-names: Abid
36given-names: Ali
37- family-names: Khan
38given-names: Dawood
39- family-names: Alfozan
40given-names: Abdulrahman
41- family-names: Zou
42given-names: James
43doi: 10.48550/arXiv.1906.02569
44date-released: 2019-06-06
45url: https://arxiv.org/abs/1906.02569
46