Update new bbs coords using imaug functions

[alice17]

No description provided.

[bkanaki]

Hey, I tried to use your patch. Did you ever encounter this error:

File "C:\Users\bhargav\AppData\Local\Continuum\miniconda3\envs\tf110gpu\lib\site-packages\tensorflow\python\keras\utils\data_utils.py", line 607, in get inputs = self.queue.get(block=True).get() File "C:\Users\bhargav\AppData\Local\Continuum\miniconda3\envs\tf110gpu\lib\multiprocessing\pool.py", line 644, in get raise self._value File "C:\Users\bhargav\AppData\Local\Continuum\miniconda3\envs\tf110gpu\lib\multiprocessing\pool.py", line 119, in worker result = (True, func(*args, **kwds)) File "C:\Users\bhargav\AppData\Local\Continuum\miniconda3\envs\tf110gpu\lib\site-packages\tensorflow\python\keras\utils\data_utils.py", line 432, in get_index return _SHARED_SEQUENCES[uid][i] File "C:\Users\bhargav\Documents\code_base\ped_det\model\preprocessor.py", line 149, in __getitem__ img, all_objs = self.aug_image(train_instance, jitter=self.jitter) File "C:\Users\bhargav\Documents\code_base\ped_det\model\preprocessor.py", line 243, in aug_image bbs_aug = seq_det.augment_bounding_boxes([bbs])[0] File "C:\Users\bhargav\AppData\Local\Continuum\miniconda3\envs\tf110gpu\lib\site-packages\imgaug\augmenters\meta.py", line 686, in augment_bounding_boxes y2=y2 File "C:\Users\bhargav\AppData\Local\Continuum\miniconda3\envs\tf110gpu\lib\site-packages\imgaug\imgaug.py", line 1524, in copy label=self.label if label is None else label File "C:\Users\bhargav\AppData\Local\Continuum\miniconda3\envs\tf110gpu\lib\site-packages\imgaug\imgaug.py", line 1259, in __init__ do_assert(x2 > x1) File "C:\Users\bhargav\AppData\Local\Continuum\miniconda3\envs\tf110gpu\lib\site-packages\imgaug\imgaug.py", line 678, in do_assert raise AssertionError(str(message)) AssertionError: Assertion failed.

[alice17]

I think something is wrong with the coordinates of your bounding boxes: x2 must be greater than x1.

[bkanaki]

I understand that from the error message. However, I am using coco 2017 dataset and I don't think there is any error in the bounding boxes for coco dataset as far as I know. In fact, I subtracted 1 and added 1 from min and max values respectively. I still get the same error, after few epochs!

[bkanaki]

I am not sure what creates this error. Here is the list of bounding boxes when this error is raised:

BoundingBoxesOnImage([
    BoundingBox(x1=509.0000, y1=63.0000, x2=628.0000, y2=252.0000, label=None), 
    BoundingBox(x1=382.0000, y1=119.0000, x2=564.0000, y2=254.0000, label=None), 
    BoundingBox(x1=467.0000, y1=44.0000, x2=498.0000, y2=123.0000, label=None), 
    BoundingBox(x1=579.0000, y1=22.0000, x2=640.0000, y2=129.0000, label=None), 
    BoundingBox(x1=188.0000, y1=52.0000, x2=351.0000, y2=244.0000, label=None), 
    BoundingBox(x1=352.0000, y1=18.0000, x2=406.0000, y2=150.0000, label=None), 
    BoundingBox(x1=288.0000, y1=27.0000, x2=344.0000, y2=149.0000, label=None), 
    BoundingBox(x1=228.0000, y1=18.0000, x2=274.0000, y2=142.0000, label=None), 
    BoundingBox(x1=180.0000, y1=23.0000, x2=227.0000, y2=67.0000, label=None), 
    BoundingBox(x1=126.0000, y1=20.0000, x2=185.0000, y2=149.0000, label=None), 
    BoundingBox(x1=52.0000, y1=18.0000, x2=128.0000, y2=150.0000, label=None), 
    BoundingBox(x1=3.0000, y1=20.0000, x2=61.0000, y2=138.0000, label=None), 
    BoundingBox(x1=392.0000, y1=4.0000, x2=437.0000, y2=43.0000, label=None), 
    BoundingBox(x1=28.0000, y1=5.0000, x2=397.0000, y2=47.0000, label=None
)], shape=(278, 640, 3))

[rodrigo2019]

@alice17 and @bkanaki I did a similar implementation and it looks working fine to me.
check here

[bkanaki]

Would the KeyPoints v.s. BoundingBox create an error? Theoretically, it shouldn't if the implementation for both are very similar in terms of sanity check.