YOLO v9 and TIFF files

Support
,
1

Hi there!

I am using YOLO v9 for object detection from sentinel 2 tiff files. The thing is I’m stuck at 2 things:

1)Should I convert my tiff to png/ jpg? But then I would lose my spectral signature and I think my object is not so visible in image so i may need spectral reflectance.

2)How to annotate and create bounding boxes for my tiff file if I don’t convert this into png?

Below is the link that Im trying to do in YOLO.

Thank you for the support.

2

  1. Should I convert my tiff to png/ jpg? But then I would lose my spectral signature and I think my object is not so visible in image so i may need spectral reflectance.

Have you tested to see how good/bad the conversion is? Additionally, have you looked to see if there are any tools that will combine the spectral reflectance channel with the base RGB channels?

  1. How to annotate and create bounding boxes for my tiff file if I don’t convert this into png?

There are numerous annotation solutions on the market, and a quick web search like “image annotation TIFF” will help you find what you need. CVAT was the first one to come up for me, and there are numerous other platforms like Encord and Label Studio, however anything browser based might be tricky. You could also use the OpenCV or Pillow Python libraries to directly annotation images, but that might be a bit more involved. There’s also ImageJ, which is more geared to microscopy data, but has tools for annotating images of nearly any format type.