zcablii / SARDet_100K

Offical implementation of MSFA and release of SARDet_100K dataset for Large-Scale Synthetic Aperture Radar (SAR) Object Detection
Other
328 stars 25 forks source link

training data set #19

Closed paster489 closed 3 months ago

paster489 commented 3 months ago

Dear Yuxuam,

i want to use your training set for yolo v9 training.

In the annotation train.json file your BBox are not matched to object in image. For example for the image file 0000026.jpg the BBox is: "images": [ { "file_name": "0000026.jpg", "height": 256, "width": 256, "id": 7622 }
],

"annotations": [

    {
        "area": 829.0,
        "iscrowd": 0,
        "image_id": 3086,
        "bbox": [
            642.0,
            11.0,
            65.0,
            39.0
        ],
        "category_id": 0,
        "id": 7622,
        "ignore": 0,
        "segmentation": []
    }
],

which is out of image size 256 x 256. i saw that 0000026.jpg is tile of "image_id": 3086 which include additional ~28 tiles. how to merge these tiles to update BBox of 0000026.jpg ? Or maybe you have another method to set correctly BBox of 0000026.jpg ?

thanks you for the help?

zcablii commented 3 months ago

It seems there might be some misunderstanding regarding the annotation interpretation. Our annotations strictly follow the COCO format. For instance, the image 0000026.jpg has an "id" of 7622, which corresponds to the "image_id" in the "annotations" section. Therefore, you need to look for annotations where "image_id" is 7622. For more details, u can google "coco dataset annotation interpretation".

paster489 commented 3 months ago

thank you

On Mon, Jun 3, 2024 at 4:45 PM Yuxuan Li @.***> wrote:

It seems there might be some misunderstanding regarding the annotation interpretation. Our annotations strictly follow the COCO format. For instance, the image 0000026.jpg has an "id" of 7622, which corresponds to the "image_id" in the "annotations" section. Therefore, you need to look for annotations where "image_id" is 7622.

— Reply to this email directly, view it on GitHub https://github.com/zcablii/SARDet_100K/issues/19#issuecomment-2145247040, or unsubscribe https://github.com/notifications/unsubscribe-auth/AO3HQPSP5MDPWRKGND6BJZ3ZFRXRPAVCNFSM6AAAAABIV6XA4KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCNBVGI2DOMBUGA . You are receiving this because you authored the thread.Message ID: @.***>