search

Jordan-Pierce / Marine-Debris

0 stars 0 forks source link

Set up #1

Open Jordan-Pierce opened 2 weeks ago

Jordan-Pierce commented 2 weeks ago

@rosswin @abatt34 @kirakaplann

I'm going to start populating this repo with some things including:

I think it would be useful to communicate here in the future.

Jordan-Pierce commented 2 weeks ago

Okay, I added some initial set up stuff. I'm still in the process of trying to download a month's data from the Google drive.

Jordan-Pierce commented 1 week ago

@rosswin @abatt34 @kirakaplann

Made progress on downloading the data from the GDrive using the API. Unfortunately this is something that can really only be done from NOAA? So, I would say wrt to playing around with the code, it's best to manually download a month of data to test. Recommend using April 2024, as that's what Kira is using.

Will share updates.

abatt34 commented 1 week ago

@Jordan-Pierce @rosswin @kirakaplann Played around w/ getting a raster dataset set up but am getting same errors that I know Kira is. seems like our JPEGs need to be georeferenced in order to use the "GeoDataset" type. I know I've done this within GIS platforms, but not in python. Thoughts? would we use GDAL? Anyone done this before? is this necessary?

Jordan-Pierce commented 1 week ago

Will check it out tomorrow. The JPEGs are georeferenced, but it could be stored in the header in a way that torchgeo isn't expecting

rosswin commented 1 week ago

Hey all,I haven’t taken a look yet, but usually these issues are due to missing georeference files. JPEGs don’t have a way to store internal georeference information (only the TIFF format can do that). Each image should have one, but possibly more, sidecar files such as .JGW or .JPG.AUX.XML. So make sure the folder that holds your JPEGs also has the matching sidecar files. If that doesn’t solve it then attach one of the troublesome images to this email and we should be able to solve that quickly. -RossOn Jul 7, 2024, at 9:41 PM, Jordan Pierce @.***> wrote: Will check it out tomorrow. The JPEGs are georeferenced, but it could be stored in the header in a way that torchgeo isn't expecting

—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you were mentioned.Message ID: @.***>

Jordan-Pierce commented 1 week ago

@rosswin

Yep you're right. The files are JPEGs but do not have side-car files. I saw Lat / Lon in the EXIF, but they're just geotagged, not georeferenced.

Below I have the annotations imported and converted as points (blue), and the images imported using the GeoTaggedPhotosToPoints tool.

Deer_Island_Test

So, we need them to provide us with the side-car files for each JPEG image, or, more difficult, provide us with a GeoTIFF version of each image?

abatt34 commented 1 week ago

seems like that's probably the best bet. Should we write an email? do you think they 100% have those files or should I research ways to do it ourselves?

rosswin commented 1 week ago

Without those sidecars we can’t get to true georeferencing. No way around it… gotta ask for those files. Once we have the sidecars we could convert the images to Tifs and embed the sidecar files into the image headers. Then we don’t have to worry about this. On Jul 8, 2024, at 2:51 PM, Allie Battista @.***> wrote: seems like that's probably the best bet. Should we write an email? do you think they 100% have those files or should I research ways to do it ourselves?

—Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you were mentioned.Message ID: @.***>