|
10th September 2019, 11:05 AM | #1 | |
EAAF Staff
Join Date: Nov 2004
Location: Upstate New York, USA
Posts: 898
|
Quote:
|
|
11th September 2019, 03:09 AM | #2 |
Member
Join Date: Dec 2004
Location: Ann Arbor, MI
Posts: 5,503
|
As far as I could understand, this may be just a training exercise for machine image recognition. Kind of testing a simplified version of facial recognition. Machine reading of CT and MRI scans is under very active development. A very simple version is machine reading of ECGs, and is already widely used.
Kris, with their complexity of forms, are very handy model. |
11th September 2019, 01:35 PM | #3 |
Member
Join Date: Sep 2005
Location: Singapore
Posts: 308
|
I work with machine learning for classifying land cover types (forest, grassland etc) on satellite imagery to better under changes over time, like deforestation or development. This particular paper is using edge detection outputs to classify keris types. As such this will broadly allow it to classify broad groups that can be distinguishable with 2-D profile (silhouette) information. As Ariel says, this an interesting academic exercise but certainly can't add to keris understanding which is clearly far more subtle than profile ratios.
|
17th September 2019, 07:42 PM | #4 |
Member
Join Date: Jan 2005
Posts: 478
|
This is cool stuff, at least to a komputer wienie like me. It seems to be a new take on algorithms to help identify an object. With a database of only 10 comparators it is a little weak. However I thing what we have here is someone has an interest in the Keris and needed to get a paper published. So wallah. Alan I don't think you have anything to worry about in the foreseeable future.
|
17th September 2019, 08:53 PM | #5 | |
Keris forum moderator
Join Date: Aug 2006
Location: Nova Scotia
Posts: 7,047
|
Quote:
|
|
18th September 2019, 02:51 PM | #6 | |
Member
Join Date: Jan 2005
Posts: 478
|
Quote:
|
|
2nd April 2020, 04:57 AM | #7 |
Member
Join Date: Jan 2018
Location: Sydney, Australia
Posts: 277
|
Haha this is definitely one of the more niche applications of ai/machine learning I've seen. As many have said, don't know if it helps us as students of the keris, but it's a fun machine learning exercise that can certainly be advanced.
You could probably use some variant or less sophisticated version of a machine learning model used in geospatial applications. If it can recognise topography (depth, height, etc.), then it will be able to distinguish the full array of ricikan that we need to be able to train the machine to make a prediction about what dhapur a keris is. Imagine being able to take a photo of your keris, uploading it to Keris Warung Kopi, and our resident machine can tell you in seconds. We might put Alan out of a job! |
2nd April 2020, 11:18 AM | #8 |
Member
Join Date: Sep 2014
Location: Austria
Posts: 1,882
|
Are you guys getting bored or what?!
|
|
|