Fitzwilliam Museum: Difference between revisions

From Computer Laboratory Group Design Projects
Jump to navigationJump to search
No edit summary
No edit summary
Line 1: Line 1:
Contact: Joanne Vine <jrv31@cam.ac.uk>
Contact: Joanne Vine <jrv31@cam.ac.uk>
Daniel Pett <dejp3@cam.ac.uk>


Potential client: Daniel Pett <dejp3@cam.ac.uk> says I’m interested in quite a few things based around collections and data:


1) Machine learning and computer vision - the work my former colleague, Harrison Pim is doing at the Wellcome is along the lines of where I want to go eg https://twitter.com/hmpim/status/1052961056727990272 Very python driven research or R.


I’m interested in quite a few things based around collections and data:
Feedback: Harrison Pim's work is probably more the kind of thing that we might get a Master's student to do. Undergraduates could probably have a good hack at it, but computer resources are higher than they usually have, it can often take longer to negotiate access to labelled data than the span of the project, and it's hard to divide this kind of work among a team - as you say, usually just one person at a Python console (followed by waiting three days for the network to be trained).


1) Machine learning and computer vision - the work my former colleague, Harrison Pim is doing at the Wellcome is along the lines of where I want to go eg https://twitter.com/hmpim/status/1052961056727990272 Very python driven research or R.
2) Predictive modelling of exhibition attendance
2) Predictive modelling of exhibition attendance
Feedback: Predictive modelling of exhibition attendance would have results largely driven by the quality of your dataset. What kind of real-time data do you have access to, and is there a substantial amount of historical data for building the predictive model?
3) Data science stuff around social media and reviews data from sources like Tripadvisor
3) Data science stuff around social media and reviews data from sources like Tripadvisor
4) 3D modelling approaches
5) Visualisation of any sort
6) expansion of my crowdsourcing project https://crowdsourced.micropasts.org


1) Harrison Pim's work is probably more the kind of thing that we might get a Master's student to do. Undergraduates could probably have a good hack at it, but computer resources are higher than they usually have, it can often take longer to negotiate access to labelled data than the span of the project, and it's hard to divide this kind of work among a team - as you say, usually just one person at a Python console (followed by waiting three days for the network to be trained).
Feedback: It's an interesting idea to work with Tripadvisor. Do Cambridge museums pick up a lot of reviews there? Do you happen to know anything about their APIs and terms of service for data mining research?


2) Predictive modelling of exhibition attendance would have results largely driven by the quality of your dataset. What kind of real-time data do you have access to, and is there a substantial amount of historical data for building the predictive model?
4) 3D modelling approaches


3)  It's an interesting idea to work with Tripadvisor. Do Cambridge museums pick up a lot of reviews there? Do you happen to know anything about their APIs and terms of service for data mining research?
Feedback: I presume you mean 3D models of collection objects. Do you have a laser scanner that could be used for the digitisation stage? We don't have one in the Computer Lab, but have collaborated with other museums in the past who used their own scanners, for example in NHM.


4) I presume you mean 3D models of collection objects. Do you have a laser scanner that could be used for the digitisation stage? We don't have one in the Computer Lab, but have collaborated with other museums in the past who used their own scanners, for example in NHM.
5) Visualisation of any sort


5) Past experience is that CS undergraduates are not that good at inventing novel visualisations. My research group does a lot of visualisation research, but for undergrads, we would probably have to specify exactly what we want their system to look like.
Feedback: Past experience is that CS undergraduates are not that good at inventing novel visualisations. My research group does a lot of visualisation research, but for undergrads, we would probably have to specify exactly what we want their system to look like.
 
6) That's a nice project. Which of the datasets have attracted most labelling input? Is this in a form that validation and reliability analysis might be useful?


6) expansion of my crowdsourcing project https://crowdsourced.micropasts.org


Feedback: That's a nice project. Which of the datasets have attracted most labelling input? Is this in a form that validation and reliability analysis might be useful?


Original introduction:


After meeting with Andrea Kells, "... feels that there might be scope for the Museum to offer one or more projects – for example around digital curation, monitoring visitor interaction with exhibits/visitor flow, using social media and other sources to collate and analyse visitor feedback etc."
After meeting with Andrea Kells, "... feels that there might be scope for the Museum to offer one or more projects – for example around digital curation, monitoring visitor interaction with exhibits/visitor flow, using social media and other sources to collate and analyse visitor feedback etc."

Revision as of 14:38, 30 October 2018

Contact: Joanne Vine <jrv31@cam.ac.uk>

Potential client: Daniel Pett <dejp3@cam.ac.uk> says I’m interested in quite a few things based around collections and data:

1) Machine learning and computer vision - the work my former colleague, Harrison Pim is doing at the Wellcome is along the lines of where I want to go eg https://twitter.com/hmpim/status/1052961056727990272 Very python driven research or R.

Feedback: Harrison Pim's work is probably more the kind of thing that we might get a Master's student to do. Undergraduates could probably have a good hack at it, but computer resources are higher than they usually have, it can often take longer to negotiate access to labelled data than the span of the project, and it's hard to divide this kind of work among a team - as you say, usually just one person at a Python console (followed by waiting three days for the network to be trained).

2) Predictive modelling of exhibition attendance

Feedback: Predictive modelling of exhibition attendance would have results largely driven by the quality of your dataset. What kind of real-time data do you have access to, and is there a substantial amount of historical data for building the predictive model?

3) Data science stuff around social media and reviews data from sources like Tripadvisor

Feedback: It's an interesting idea to work with Tripadvisor. Do Cambridge museums pick up a lot of reviews there? Do you happen to know anything about their APIs and terms of service for data mining research?

4) 3D modelling approaches

Feedback: I presume you mean 3D models of collection objects. Do you have a laser scanner that could be used for the digitisation stage? We don't have one in the Computer Lab, but have collaborated with other museums in the past who used their own scanners, for example in NHM.

5) Visualisation of any sort

Feedback: Past experience is that CS undergraduates are not that good at inventing novel visualisations. My research group does a lot of visualisation research, but for undergrads, we would probably have to specify exactly what we want their system to look like.

6) expansion of my crowdsourcing project https://crowdsourced.micropasts.org

Feedback: That's a nice project. Which of the datasets have attracted most labelling input? Is this in a form that validation and reliability analysis might be useful?

Original introduction:

After meeting with Andrea Kells, "... feels that there might be scope for the Museum to offer one or more projects – for example around digital curation, monitoring visitor interaction with exhibits/visitor flow, using social media and other sources to collate and analyse visitor feedback etc."

Previous projects (see also Hamilton Kerr Institute, Cambridge Museums):

  • Next-Generation Museum Guide - in 2008 and 2006
  • Smart Poster Picker - in 2008