Can we use AIAA for non-medical 2D images

As it took much more time to manually draw a bounding box around the object of interest for 2D png/jpg image, can we use AIAA to reduce the time over there?

yes… you should be able to solve this problem using AIAA.
you can compute your own bounding box as part of post-transform and add it in aiaa config. AIAA will forward both result mask and json results back to user/client

For example (this can be your own transform which you can bring into AIAA (bring your own transforms):

Add any textual results to transform_ctx (with key: ‘result_params’)

class AddBoundaryBox(Transformer):
    def __init__(self, image_field='image', label_field='model', params='result_params', box='box'):
        self._image_field = image_field
        self._label_field = label_field
        self._params = params
        self._box = box
    def transform(self, transform_ctx: TransformContext):
        image = transform_ctx.get_image(self._image_field)
        label = transform_ctx.get_image(self._label_field)

        # Your logic to compute bounding box for input image+label 
        box = my_box(image, label)

        params = transform_ctx.get(self._params, {})
        params[self._box] = box
        transform_ctx[self._params] = params
        return transform_ctx

And use this kind of transform in your aiaa-config under post transforms


hi,I have a 2D detection task in Abdominal lesions. There are similar questions that how to draw bounding box in the process of inference. Besides, can you tell me how to show the bounding box in 3D-slicer by AIAA plugin?