search

DocF / multispectral-object-detection

Multispectral Object Detection with Yolov5 and Transformer
GNU Affero General Public License v3.0
272 stars 55 forks source link

FLIR Google Drive Link points to random eBay listing #63

Open ntinou opened 8 months ago

ntinou commented 8 months ago

Hello,

I tried to download the aligned FLIR dataset using the link in the README but it points to a random eBay listing instead of a google drive folder. Is there any chance that you can provide the correct link?

EST09 commented 8 months ago

I think I used the aligned one from here: https://github.com/CalayZhou/Multispectral-Pedestrian-Detection-Resource

ntinou commented 8 months ago

Hi, Thanks for the assist, I'm posting a link to the folder itself here, for future reference to anyone also trying the same dataset.

@EST09 Do you perhaps remember if you used the align txts as they are in the zip, because I'm looking at the example FLIR_aligned.yaml file and it's looking for separate rgb and ir txts while the zip doesn't make that distinction.

Thanks again!

EST09 commented 8 months ago

If I remember rightly I made my own duplicate .txt file for the equivalent files in the other modality to point train_ir, val_ir to

ntinou commented 8 months ago

That makes sense, I'll follow in your tracks.

Thank you for your help!

EST09 commented 8 months ago

No worries! If you get far with this, could I ask if you ever manage to get it to run with images from just one modality? That's where I'm struggling atm but I don't think that's what the model is designed for tbh @DocF

ntinou commented 8 months ago

That's what I'm going to try to do as well. If I have any luck I'll be back with an update.

EST09 commented 8 months ago

Thanks! That would be great if you do make any headway!

mayone23 commented 3 months ago

Could you please help me in setting the working dataset directory? I have downloaded from the above shared link (Thanks @EST09 @ntinou for the dataset). I don't understand how to set the dataset directory and where should I keep the dataset and how the annotation files are used. Because there are three folders for images named as Annotations, JpegImages and Annotated images. Among these which one will be going to use? Thanks.