Spaces:

Voxel51

/

DataCentricVisualAIChallenge

like 12

Running

Hi thanks for taking part in the competition.

The HF space for the competition has a lot of details you may be interested in and which will clarify your questions: https://huggingface.co/spaces/Voxel51/DataCentricVisualAIChallenge

I'll answer your questions below:

Which machines are we supposed to be using? For the data curation, training and inference?

You can use whatever machine you want while you are developing a submission. The end result will be the submission of a data curation script and a config file for model training. See the repo here for more details, and the submission guideline page of the hugging face space. When you submit your data curation script, we will train the model and evaluate.

Once a model is trained (with the data curation in place), where should it be uploaded?

You don't submit a model, you just submit your data curation script and config for model training. we'll evaluate and put your score on the leaderboard

For the baseline, it would be great if you can post the metrics value on the public eval set using yolov8m finetuned on the training set (with NO data curation)

Thanks for the suggestion

Can we have some instructions on the submission procedure please?

Again, please carefully review all the pages in the hugging face space and the example submission repo. I've linked to both here

HF Space: https://huggingface.co/spaces/Voxel51/DataCentricVisualAIChallenge

Example repo: https://github.com/harpreetsahota204/DCVAI-Challenge-Example-Submission/tree/main