geo-deep-learning
geo-deep-learning copied to clipboard
NRCan
Deep learning applied to georeferenced datasets
fixes #215 Implements a new script for tiling using the solaris library and replacing the use of patches saved as numpy arrays in HDF5 files. The outputted patches are .tif...
In the production inference process, there is a step where we check the presence or not of the asset in the data repository to avoid downloading it in HPC every...
Our worldview images generally have a histogram skewed toward lower values, with a low contrast. Latest tests by @valhassan show that a light enhancement with the CLAHE algorithm applied on...
[222-stac-item-input](https://github.com/NRCan/geo-deep-learning/tree/222-stac-item-input) branch have redundant function that load the checkpoint each time. - [ckpt_is_compatible(checkpoint)](https://github.com/NRCan/geo-deep-learning/blob/222-stac-item-input/inference_segmentation.py#L237) - [checkpoint_converter](https://github.com/NRCan/geo-deep-learning/blob/222-stac-item-input/inference_segmentation.py#L238) - [read_checkpoint](https://github.com/NRCan/geo-deep-learning/blob/222-stac-item-input/inference_segmentation.py#L239) - [InferenceTask.load_from_checkpoint(checkpoint)](https://github.com/NRCan/geo-deep-learning/blob/222-stac-item-input/inference_segmentation.py#L297)
 Pynvml dependency is used in torch an error arises in a new environment without pynvml installed _Originally posted by @valhassan in https://github.com/NRCan/geo-deep-learning/pull/344#discussion_r986328985_
Though new conda environments created with GDL's [environment.yml](https://github.com/NRCan/geo-deep-learning/blob/develop/environment.yml) will install torchmetrics > 0.9 (see screenshot below), this library **and** version requirement is not explicit in this file since torchmetrics is...
The extraction of our models is done on the whole extent of the images without considering the NODATA. Frequently we get false positives - especially from water - in the...
This PR is not meant to be merged, only create show diff with develop. Discussions ongoing as to how this branch should be merged. FYI, around 4000 of 6265 added...
See example here: https://github.com/remtav/geo-deep-learning/blob/222-stac-item-input/config/inference/production.yaml#L24
[Kornia](https://kornia.readthedocs.io/en/latest/index.html) offers great transforms for data augmentation. GDL could replace its in-house augmentations from [augmentation.py](https://github.com/NRCan/geo-deep-learning/blob/develop/utils/augmentation.py) for Kornia's ones. Initially, the problem was that torchvision's transforms were based on PIL and...