Woodscape Fisheye Object Detection Challenge for Autonomous Driving | CVPR 2022 OmniCV Workshop Forum

Go back to competition Back to thread list Post in this thread

> AttributeError: 'list' object has no attribute 'astype'

Hi ALL:
When I first commit , I found this Error as below :

WARNING: Your kernel does not support swap limit capabilities or the cgroup is not mounted. Memory limited without swap.
Traceback (most recent call last):
File "/tmp/codalab/tmpQewVjR/run/program/evaluate.py", line 89, in <module>
score_task = evaluate_tasks[task](pred, gt, settings[task])
File "/tmp/codalab/tmpQewVjR/run/program/evaluate_tasks.py", line 35, in evaluate_object_detection
ious = bbox_overlaps(bboxes_pred, bboxes_gt)
File "/tmp/codalab/tmpQewVjR/run/program/utils.py", line 99, in bbox_overlaps
bboxes1 = bboxes1.astype(np.float32)
AttributeError: 'list' object has no attribute 'astype'

Then I have Debug the Eval Code , I find if the gt has the class eg: 3 , and the predict do not has this class, then the program will get above error!
So Please Check the Code!
Thanks!

Posted by: caic-ailab @ April 27, 2022, 12:49 a.m.

May be you can fix this bug in the Eval Code like this:

In evaluate_tasks.py file , before modify the line 28,29 is like this:

bboxes_pred = pred[i][label_gt] if label_gt in pred[i] else []
bboxes_gt = gt[i][label_gt] if label_gt in gt[i] else []

we can modify it as below:

bboxes_pred = pred[i][label_gt] if label_gt in pred[i] else np.array([])
bboxes_gt = gt[i][label_gt] if label_gt in gt[i] else np.array([])

the this bug will fixed!

GoodLucky!

Posted by: caic-ailab @ April 27, 2022, 12:56 a.m.

Dear caic-ailab,

We're sorry that your submission resulted in an error. Thanks for reporting this.
We encountered another error and this will be fixed soon.

Regards,
Saravanabalagi.

Posted by: saravanabalagi @ April 27, 2022, 1:17 a.m.

Hi all,
This issue has been fixed and you should be able to submit without errors.
Cheers,
Saravanabalagi.

Posted by: saravanabalagi @ April 27, 2022, 1:53 a.m.

Sorry to trouble you, i have found that the label only has 3 class, but you want 5 class out.
Please Check!

Posted by: caic-ailab @ April 27, 2022, 9:56 a.m.

Dear caic-ailab,
Thanks for bringing this to our notice. We’ll get back to you soon. I appreciate your patience.
Regards,
Saravanabalagi.

Posted by: saravanabalagi @ April 29, 2022, 4:29 p.m.

Dear caic-ailab,
The 2D bounding box annotations for the training set have been updated in the official Woodscape repository to include class 4 and 5 (traffic_light and traffic_sign), which were previously missing due to an error. We kindly request you to re-download the ground truth labels for the training set (the links are provided in Participate > Get Data section).
Regards,
Saravanabalagi.

Posted by: saravanabalagi @ May 6, 2022, 2:22 p.m.
Post in this thread