Hacker News new | past | comments | ask | show | jobs | submit login

What is really interesting is that the main innovation is never mentioned in the BBC article. The academic they quote, Dave Cowley, is an author on "Using deep neural networks on airborne laser scanning data: Results from a case study of semi‐automatic mapping of archaeological topography on Arran, Scotland" from 11/2018 [0]. The "new 3D technology" that is becoming more "widely available" and allows for "rapid discovery" is not LIDAR (which is very old) but Deep Learning as applied to LIDAR. It's interesting b/c the BBC doesn't mention "Deep Learning" in the article. LIDAR is old enough that it was launched into space as early as 1971. Deep learning is most popular on images, but is becoming more popular recently on less structured data like the unordered collection of points in xD (x >= 3), so this is the new part.

(I will also say that in this case, the article makes images out of the LIDAR data and runs DL on that, but this is not mandatory. Checkout Pointet++ if interested. I just wanted to draw attention to the fact that the "cutting edge" part was not collecting data but analyzing it.)

[0] https://onlinelibrary.wiley.com/doi/full/10.1002/arp.1731




Do you happen to know if anyone out there is using DL techniques for identifying and most importantly for categorizing ancient pottery? I’m asking because until now most of that identifying and categorization has been of course made only “manually”, by archeologists themselves, with, imho, different levels of “success”. On top of genuine human errors (i.e. no-one can visually remember and compare all the pottery from a certain geographical region) there is also the added geo-political factor that must be taken into consideration, i.e. an archeologist from a specific country won’t categorize that easily a certain piece of pottery as belonging to a neighboring country/culture and thus a potential “enemy”. I’m thinking that further automatizing this process will get rid of human error.


p.s. I think the LIDAR data is here: https://remotesensingdata.gov.scot/collections , not on Canmore. It seems to be about a gigabyte total.


Thank you for this!




Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact

Search: