Could you please share the training method ?
First, great job.
I understand it is a YOLOv8 based model. The output misses some parts, so I was wondering which dataset it was trained against ? news-navigator ?
Thank you! Your feedback is appreciated. Yes, it does miss a lot, and mislabels some objects as well, we tried our best to fix them and are welcoming any feedback to make it better.
This is a demo to our solution to https://www.kaggle.com/competitions/dlsprint2/overview,
dataset it was trained on: https://www.kaggle.com/competitions/dlsprint2/data?select=badlad
you can find the whole solution(including training, data augmentation, inference scripts, all trained weights) here: https://www.kaggle.com/competitions/dlsprint2/discussion/432099
We will upload the paper explaining the whole procedure soon. We suggest you go through the presentation slide to understand our solution better in the mean time.
Thank you so much. Appreciate it.
Thank you very much. Please indicate the paper as soon as it's ready.
Are your models AGPL licensed as well?
Link to the paper: https://arxiv.org/abs/2309.00848
And no, our models are not AGPL licensed.