Closed wwdlh closed 2 months ago
hi @wwdlh ,
Thanks for your interest in our work. We have not used V1.0-mini for training and evaluation since its data is not enough for reasonable convergence. However, you can evaluate our released model on it by replacing the data-related paths as the bbox-wise gt is generated automatically. Welcome to post your results here!
The amount of data in v1.0-trainval is too large