Try to make sure all Image IDs are in the submission file. Say there is no detection for the Image ID(Most likely the case) that is missing, you could just fill corresponding bounding boxes with zeros per class.
This is strange, cause if 80% of dataset is training data, to include 80% of data with either perfect score or zero score would definitely affect the final submission outcome
Yeah, It seems there might be some limitations on the number of requests for accessing the data on the Drive, which could be causing the issue. In such cases, it's usually best to wait until tomorrow and try again. The system may reset the access limits after some time. I faced also this situation.
Try to make sure all Image IDs are in the submission file. Say there is no detection for the Image ID(Most likely the case) that is missing, you could just fill corresponding bounding boxes with zeros per class as said @crossentropy.
This is what worked for me too, but i think it reduce the score.
Please I need help. Are we to submit the csv of our training and testing output or the entire notebook? I am getting this error "Missing columns Image_ID on the submission file" any help ?
Try to make sure all Image IDs are in the submission file. Say there is no detection for the Image ID(Most likely the case) that is missing, you could just fill corresponding bounding boxes with zeros per class as said @crossentropy.
This is one way but it reduces the score.
You can also do a recursive function that reduces the confidence level for images with no detection until it lands a detection. However it also reduces the score. I am still looking for another solution to solve that. Maybe retraining the model might solve that
Try to make sure all Image IDs are in the submission file. Say there is no detection for the Image ID(Most likely the case) that is missing, you could just fill corresponding bounding boxes with zeros per class.
This is what worked for me.
This is strange, cause if 80% of dataset is training data, to include 80% of data with either perfect score or zero score would definitely affect the final submission outcome
That's exactly the issue I'm currently dealing with. If anyone has found an alternative solution, it would be greatly appreciated.
Hello Peniel, how were you able to access the data from the drive. I have tried accessing it ,but it doesn't work mostlikely due to usage permisions.
Yeah, It seems there might be some limitations on the number of requests for accessing the data on the Drive, which could be causing the issue. In such cases, it's usually best to wait until tomorrow and try again. The system may reset the access limits after some time. I faced also this situation.
Alright Thanks. I was able to download from kaggle via the Api
You're welcome
how were you able to rectify this
Error. Image ID id_1h4quq.jpg not found in submission file
Hello brother ,
Try to make sure all Image IDs are in the submission file. Say there is no detection for the Image ID(Most likely the case) that is missing, you could just fill corresponding bounding boxes with zeros per class as said @crossentropy.
This is what worked for me too, but i think it reduce the score.
Please I need help. Are we to submit the csv of our training and testing output or the entire notebook? I am getting this error "Missing columns Image_ID on the submission file" any help ?
Just the submission csv file.
Try to make sure all Image IDs are in the submission file. Say there is no detection for the Image ID(Most likely the case) that is missing, you could just fill corresponding bounding boxes with zeros per class as said @crossentropy.
This is one way but it reduces the score.
You can also do a recursive function that reduces the confidence level for images with no detection until it lands a detection. However it also reduces the score. I am still looking for another solution to solve that. Maybe retraining the model might solve that
You could also try reducing your filter threshold to permit all predictions to be listed in the submission file regardless of its performance