Preliminary Test Phase and Guidelines

Dear Participants,

Thank you for your interest and active participation in our challenge!

As of August 17th, we are thrilled to have over 240 participants from across the globe. Tomorrow, we will close the application channel, and it will remain closed until after the challenge concludes.

The preliminary test phase for both tasks will officially begin at 12:00 AM (Beijing time) on August 18th. Please take note of the following important details and guidelines:

1. Preliminary Test Overview: Each task will include 20 images across 4 categories, with 5 images in each category. Two categories will be from the training set, while the other two will introduce new categories. The ranking will be determined with a weight of 0.2 assigned to the Preliminary Test Set and 0.8 to the Final Test Set.

2. Test Duration: You will have 2 weeks to submit and test your algorithms. The deadline for the preliminary test is 24:00 (Beijing time) on August 31st.

3. Submission Guidelines: To ensure fairness and avoid over-optimization, we ask that you adhere to the following rules when submitting your algorithms:

   • Each account is allowed 5 submissions during the preliminary test phase, with a maximum of 1 submission per day.
   • If your team uses multiple accounts, please refrain from submitting similar algorithms across different accounts during both the preliminary and final stages. This rule is in place to prevent algorithm adjustments based on test set scores.
   • If your team has multiple distinct methods to solve the challenge tasks, we encourage you to submit these algorithms under different accounts. However, to maintain fairness, each team is allowed a maximum of 2 accounts for submission. Any violations of these rules will result in disqualification and forfeiture of scores.

4. Final Test Invitations: After the preliminary test, we will invite the top 20 teams from each task to participate in the final test submission. The final test stage will last only 2 days, and each team will have just 1 submission opportunity.

If you have any questions or need further assistance, please feel free to reach out to us.

Thank you once again for your participation, and best of luck in the challenge!

Best regards,
Challenge Organizing Committee

Hi!

Is there any diffierence between the Preliminary Test Phase and Docker Testing Phase in the docker file? I submit the docker image which run sucessfully in Docker Testing Phase , but got error in Preliminary Test Phase . And I could only see Import Failed, how to fix it ? Does this lead to the reason? "Validation Errors This container image has already been uploaded. Please re-activate the existing container image or upload a new version."

By the way, if the submit is the same, why cannot we just choose the algorithm from Docker Testing Phase ? And have to upload the big file again.

Dear COSAS2024 organizer,

Hi

I see that the input in Task2 includes not only the test image, but also "scanner (String) at /input/scanner.json". Could you please clarify the purpose of this scanner.json?

Many thanks

Best regards

Pengzhou

I encountered the same issue: 'Validation Errors - This container image has already been uploaded. Please re-activate the existing container image or upload a new version.'

I used Docker images that had successfully passed testing.

Best regards,

Dear winycg and hoheon0509,

I apologize for the inconvenience caused by the issues with posting Docker. I am currently seeking assistance from the Grand Challenge team, but I cannot guarantee that the problem will be resolved immediately.

As you may have noticed, a temporary solution to this issue is to post the Docker image with a different MD5 hash. Once I receive an updated solution from the Grand Challenge team, I will share it on the forum as soon as possible.

Thank you for your understanding and patience.

Best regards,

Biwen Meng

Dear 柑橘乌云,

I apologize for the confusion caused by the additional file in the test data. This file was initially intended to provide more information about each image. However, we decided not to use it because not all images have this information available in practice.

Thank you for your understanding.

Best regards,

Biwen Meng