FAQ's

Submit your questions to datageneration.abc2025@gmail.com with the subject line "7th ABC Challenge." To change team members, email addresses, or other details, email the organisers at the same address.

You can also check the Tutorial page and recorded tutorial videos as most questions were discussed there.

• Link to Tutorial Page

• Link to recorded English Tutorial

• Link to recorded Japanese Tutorial

Q1. Why is the artificial dataset restricted to 5MB?

Ans. The artificial dataset is restricted to 500MB to ensure that we can effectively evaluate the performance of your virtual data generation algorithm, rather than focusing on training a robust HAR model. We believe that this size limit is sufficient for algorithm evaluation, as it allows for the generation of data that is approximately five times larger than the actual training data. 

Q2. Is the HAR model only trained on the artificial data or evaluated as well?

Ans. The training set is (a part of the real Openpack data of Scenario 1 + virtual data), and the test set is (the rest of the real Openpack data of Scenario 1). We will change the random seed and split portion of the training/set sets during evaluation. So, be sure your algorithm can work well in various settings.

Q3. Do you generate the data by the participants’ provided code? Or evaluation is done on the uploaded data only?

Ans. We will utilize both strategies to evaluate the submitted method. So, you have to provide both the Python file of the data generation algorithm and the synthetically generated data file. 

Q4. Is there any limit to the size of the original data which will be used to train the algorithm to generate the synthetic data?

Ans. No, the size of the original data is not limited. You can use any data you want to generate the data. However, the size of the generated data is limited.

Q5.  If the size of the generated data is more than 500 MB, can we submit the result by splitting it into two different parts?

Ans. No, the maximum amount of data to be generated is 500 MB only, so you cannot submit two different files.

Q6. For training the algorithm, can we use the evaluation split of your data?

Ans. Yes, you can use the evaluation split of the provided data for the training.

Q7. Where can we submit the paper and results?

Ans. You can submit your challenge paper using CMT - Choose “Challenge” as your track and select the specific challenge you are participating in. For the results of your analysis, we will share a separate submission form.

Q8. What  is the format for paper submission?

Ans. ABC challenge papers follow the IJABC Journal format as indicated in the main ABC website. Ideally about 10 pages in journal format including references. Though IJABC has flexible length, minimum 10 pages would be reasonable to explain your approach properly with supporting related work. For the template, kindly check format from https://autocare.ai/abc2025#cfp