empty box in image

[ghost]

Hello!

I try to run experiments on my custom dataset, but the following error has occurred in here
IndexError: index 0 is out of bounds for dimension 0 with size 0

In my dataset there are images without bbox annotation, which cause the above error.
i.e., y_loc_img[0] = tensor([], device='cuda:0', size=(0, 4)) for images without bbox annotation.

What does the condition mean y_loc_img[0][0][0] < 0?

[yuantn]

Hello!

MI-AOD needs to add any bounding box to each image in the unlabeled set (the positioning is not required to be accurate). For details, please refer to Question 3 in the Custom Modifications part of FAQ and the corresponding Issue.

The condition y_loc_img[0][0][0] < 0 means that the current data batch is an unlabeled data batch, because we have set all coordinates of the bounding boxes to -1 for the unlabeled data in Lines 70-74 in epoch_based_runner.py (as described in Question 7 in the Training and Test part of FAQ and the corresponding Issues).

[ghost]

The answer to question 3 in the Custom Modifications part of FAQ is a bit uncomfortable.
The answer is only related to the condition y_loc_img[0][0][0] < 0?
If then, What if we do it this way??

1. Add below method to epoch_based_runner.py:

def _add_dataset_flag(self, X, is_unlabeled):
    or _img_meta in X['img_metas'].data[0]:
     _img_meta.update({'is_unlabeled': is_unlabeled})

and add these codes
self._add_dataset_flag(X_L, is_unlabeled=False) to line 31 in epoch_based_runner.py
self._add_dataset_flag(X_L, is_unlabeled=False) to line 60 in epoch_based_runner.py
self._add_dataset_flag(X_U, is_unlabeled=True) to line 79 in epoch_based_runner.py

2. Replace y_loc_img[0][0][0] < 0 with img_metas[0]['is_unlabeled'] in MIAOD_head.py (e.g., 479 and 565 line)

[yuantn]

The code you provided is a good alternative.
It has the same effect as the original code, both for marking and judging unlabeled data.