Releases: azavea/raster-vision-data
Integration test models for v0.12
Models for PyTorch integration tests
These are being released as part of RV 0.10
Chip classification and semantic segmentation PyTorch/fastai models for integration tests
Pre-trained model for object detection integration test
Models valid for raster vision commit 187f2dba326c2567cb16d983b7faebb84d717ee2
They will be used as a pretrained model when the integration test runs in CI to speed things up.
Object Detection Integration Test model: This model was trained using the object detection integration test experiment to passing.
Chip Classification Integration Test model: This model was trained using the chip classification integration test to almost passing.
Semantic segmentation integration test pre-trained model
This model was trained using the semantic segmentation integration test experiment (with steps=5000 and batch_size=8). It will be used as a pretrained model when the integration test runs in CI to speed things up.
Partially trained model for object detection integration test
This is a model that has been partially trained (for 2500 steps) on the object detection integration test workflow, and has an F1 score of ~0.9. If trained for another 1000 steps, it gets to an F1 score of 1.0.
cowc-potsdam prediction packages
Classification and object detection prediction packages to be used with the predict_package command.
Add JPG cowc-potsdam test data
This is to test the ability to handle non-georeferenced data.
Add cowc-potsdam classification model
This is a Resnet50 model that was pretrained on Imagenet that gets an F1 of 0.97 on test dataset.
Add cowc-potsdam object detection model and test data
Initial release of car test data and model.