Skip to content

CHANGELOG.md

Latest commit

 

History

History
224 lines (146 loc) · 7.43 KB

CHANGELOG.md

File metadata and controls

224 lines (146 loc) · 7.43 KB

Changelog

All notable changes to this project will be documented in this file.

The format is based on Keep a Changelog, and this project adheres to PEP 440 and uses Semantic Versioning.

[1.0.6] - 2025-12-11

Fixed

  • Resolves #71 - median drift in cusum was incorrect.

[1.0.5] - 2025-10-27

Added

  • Function to organize burst time series into windowed data

Changed

  • Removed view/permute in favor of more readable einops to ensure self-documenting reshaping, transposing, etc

Fixed

  • Notebook's curation of data for burst application.
  • Proper model validation for mahalanobis.py pydantic models - removes warnings during test time.
  • Update environment for gpu to utilize regex for cuda compatibility.

[1.0.4] - 2025-10-02

Fixed

  • Fixed asf search for notebook examples:
    • filters out dates with single polarization
    • allows users to select which dual polarization to use (HH+HV vs. VV+VH)
  • Baseline curation matches dist-s1.

Added

  • Add models for v0 and v1 that use 32 x 32 input size.
  • Dynamically update input_size (i.e. the size of the input image) for inference using the corresponding model attribute.
  • Expose fill value of model inference (via fill_value) that was reminant of old logit transformation and is now explicitly set and can be modified

[1.0.3] - 2025-08-25

Changed

  • Latest weights for transformer_anniversary_trained_optimized_fine and transformer_anniversary_trained_optimized the v1 datasets.

[1.0.2] - 2025-08-6

Added

  • New models from v1 dataset.

[1.0.1] - 2025-07-20

Changed

  • Load model config and obtain maximum sequence length (context length) of model.
  • Decouple loading of config and weights.

Fixed

  • Changelog action

Added

  • tests for external model loading

[1.0.0] - 2025-03-20

Added

  • Nodata interpolation for interior datapoints (i.e. within pixels collected during acquisition) which are np.nan including:
    • bilinear - an iterative bilinear-like interpolation
    • nearest - nearest neighbor interpolation
  • dependabot
  • Several models for easier testing
  • notebooks that can do both immediate and anniversary selection to establish baseline

Changed

  • Removes logit transformation from normal param estimation
  • 4 new models for testing
  • estimate_normal_params_of_logits --> estimate_normal_params
  • Reorganization of transformer metric code into loading model(s), inference, torch model, and metric.
  • Updated notebook so we only download necessary arrays.

Removed

  • Notebook for multiple burst analysis (will add back later)

[0.0.14] - 2025-03-20

Fixed

  • Version/Changelog

[0.0.13] - 2025-05-15

Added

  • Added optional compilation for CPU via torch.compile and GPU via torch_tensorrt.compile.
  • 'external' option for model_token parameter of load_transformer_model.
  • And the associated cfg and wts file paths.

Fixed

  • Fixed order of tensor operations and moving data to 'device' to reduce overhead.

[0.0.12] - 2025-03-05

Added

  • Test to ensure cpu and None interface work correctly for transformer metric.

Fixed

  • The device argument for transformer needs to be correctly passed throughout the transformer estimation process and test suite.

[0.0.11] - 2025-03-05

Added

  • Ability to control device for transformer model

[0.0.10] - 2025-02-20

Added

  • Ensures environment.yml and environment_gpu.yml are up to date with the latest dependencies.
  • Requires rasterio>=1.4.0 for merging to be consistent.

[0.0.9] - 2025-02-20

Fixed

  • Bug in reproject_arrays_to_target_crs.

Added

  • Tests for merge categorical arrays using inputs that are in different crs.

[0.0.8] - 2025-02-20

Fixed

  • Pyproject.toml kept dependencies for mpire and dill (those are now removed)

[0.0.7] - 2025-02-20

Fixed

  • Pyproject.toml and environment.yml to require python 3.12+
  • Pydantic model_config to allow arbitrary types (update for v2)

Removed

  • mpire dependency

Added

  • Test for despeckling of cropped data

[0.0.6] - 2025-01-18

  • Fixed default keyword argument for merge_categorical_arrays from mode (not possible with rasterio) to min.

[0.0.5] - 2025-01-18

Added

  • Added rio_tools for merging float arrays (including burst data such as the computed metric)
    • can be used to average over overlapping areas
    • average using the distance from exterior mask (mask that touches one of the four edges)
    • exterior mask can also be dilated to avoid problematic boundary pixels
    • categorical merging of data
  • Added nd_tools for getting exterior mask and distance from such mask (wrappers around scipy.ndimage)
  • A notebook for applying the transformer metric to a large area of interest

Removed

  • Removed torch.compile from transformer model loading

[0.0.4] - 2025-01-18

Fixed

  • Fixed transformer to ensure last pre-image always has correct index

Added

  • Arxiv link to README
  • Installation instructions
  • tqdm description for despeckling
  • Allow user to disable tqdm for despeckle

Changed

  • Renamed load_trained_transformer_model to load_transformer_model
  • Renamed inputs to reflect (possible) usage with other polarizations: vv -> copol and vh -> crosspol. The APIs don't change, just the variable name inputs.

[0.0.3] - 2025-01-16

Fixed

  • Fixed ASF calls in landslide transformer notebook - more data is becoming available so static slices do not work

Added

  • Consistent linting of DIST-S1 repositories
  • Expose estimation of the normal parameters of the logit images with high/low memory usage.
  • Added transformer test based on usage of different strategies of computation.
  • Uploaded latest model

Changed

  • Data directory structure
  • Only support python 3.12+

[0.0.2]

Fixed

  • Fixes Changelog format to ensure proper automated release
  • Fixes secrets in workflows for proper automated release

[0.0.1]

This was an initial release of the library.

This is a python library for calculating a variety of generic disturbance metrics from input OPERA RTC-S1 time-series including a transformer-based metric from Hardiman-Mostow et al., 2024. Generic land disturbances refer to any land disturbances observable with OPERA RTC-S1 including land-use changes, natural disasters, deforestation, etc. A disturbance metric is a per-pixel function that quantifies via a radiometric or statistical measures such generic land disturbances between a set of baseline images (pre-images) and a new acquisition (post-image). This library is specific to the dual-polarization OPERA RTC-S1 data and will likely need to be modified for other SAR data. The user is expected to provide/curate the co-registered baseline imagery (pre-images) and the recent acquisition (post-image) for the computation of the distmetrics.

Provides an interface for the following metrics:

  1. transformer - this is metric that uses spatiotemporal context to approximate a per-pixel probability of baseline images. This that was proposed in Hardiman-Mostow, 2024.
  2. logratio - this is not a non-negative function just a difference of pre and post images in db, but we transform into a metric by only inspecting the decrease in a given polarization
  3. Mahalanobis 1d and 2d - based on sample statistics computed in patches around each pixel.
  4. CuSum - based on the actual residuals and on a normalized time-series assuming normality.