We'd like to show people how to use RV on http://xviewdataset.org/. The first step is to train a model on a small subset of the scenes. I think we should start with a Coco pretrained model (and not an xView pretrained model) to make sure the model is actually learning something. This requires writing some scripts to convert their format into an RV workflow and label files. We will probably need to improve runtime (#242, #152, #252) before training on the full dataset. Eventually we should publish a blog / tutorial on this.
We'd like to show people how to use RV on http://xviewdataset.org/. The first step is to train a model on a small subset of the scenes. I think we should start with a Coco pretrained model (and not an xView pretrained model) to make sure the model is actually learning something. This requires writing some scripts to convert their format into an RV workflow and label files. We will probably need to improve runtime (#242, #152, #252) before training on the full dataset. Eventually we should publish a blog / tutorial on this.
This is a good blog post about the dataset https://medium.com/@dariusl/object-detection-baselines-in-overhead-imagery-with-diux-xview-c39b1852f24f