# Object detection in satellite imagery

**URL:** <https://discuss.roboflow.com/t/object-detection-in-satellite-imagery/344>\
**Category:** 🤝  Community Help\
**Created:** [April 19, 2022, 6:13pm UTC](https://discuss.roboflow.com/t/object-detection-in-satellite-imagery/344 "2022-04-19T18:13:13Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![nikschet](https://avatars.discourse-cdn.com/v4/letter/n/65b543/32.png) [@nikschet](https://discuss.roboflow.com/u/nikschet)\
**Post date:** [April 19, 2022, 6:13pm UTC](https://discuss.roboflow.com/t/object-detection-in-satellite-imagery/344/1 "2022-04-19T18:13:13Z")

</div>

Hello! I would like to perform object detection (ships, airplanes) on satellite images (mainly from sentinel, images are \*.tiff).

A. Can the system support such big image files? (around 850Mb per image) Should i crop and insert them?

Moreover

B. Can i use YOLT instead of YOLO since it is tuned for such big images?

Many many thanks in advance 🙂

---

<div class="post-metadata">

**Author:** ![Mohamed](https://yyz1.discourse-cdn.com/flex029/user_avatar/discuss.roboflow.com/mohamed/32/62_2.png) [@Mohamed](https://discuss.roboflow.com/u/Mohamed)\
**Post date:** [April 20, 2022, 2:41pm UTC](https://discuss.roboflow.com/t/object-detection-in-satellite-imagery/344/2 "2022-04-20T14:41:55Z")

</div>

Hi, I would recommend cropping and inserting the images, as you’ll likely end up resizing them when performing inference too.

You can use our [model library](https://models.roboflow.com/) with our [python package](https://docs.roboflow.com/python) export code to train a custom model. We do not yet have a pre-written notebook for YOLT but if we can export in the required annotation format, you can still train that model.

---

<div class="post-metadata">

**Author:** ![nikschet](https://avatars.discourse-cdn.com/v4/letter/n/65b543/32.png) [@nikschet](https://discuss.roboflow.com/u/nikschet)\
**Post date:** [April 21, 2022, 11:54am UTC](https://discuss.roboflow.com/t/object-detection-in-satellite-imagery/344/3 "2022-04-21T11:54:15Z")

</div>

Thank you very much for the detailed answer!

---

<div class="post-metadata">

**Author:** ![Mohamed](https://yyz1.discourse-cdn.com/flex029/user_avatar/discuss.roboflow.com/mohamed/32/62_2.png) [@Mohamed](https://discuss.roboflow.com/u/Mohamed)\
**Post date:** [April 21, 2022, 1:06pm UTC](https://discuss.roboflow.com/t/object-detection-in-satellite-imagery/344/4 "2022-04-21T13:06:05Z")

</div>

You’re welcome!

---

<div class="post-metadata">

**Author:** ![AdnanF](https://avatars.discourse-cdn.com/v4/letter/a/3ab097/32.png) [@AdnanF](https://discuss.roboflow.com/u/AdnanF)\
**Post date:** [April 13, 2024, 3:50am UTC](https://discuss.roboflow.com/t/object-detection-in-satellite-imagery/344/5 "2024-04-13T03:50:58Z")

</div>

Hi, I have some brief experience with this kind of a problem. I am assuming you have GeoTIFFs and those will be too large. Here is a very handy tool on Github that I found that will do the cropping (and if it has labels, then that is taken care of too): [https://github.com/frhass/yolo\_dataprep](https://github.com/frhass/yolo_dataprep)

Go with using 640px x 640px as per the default YOLO usage.

Hope it helps.