v0.3.0 (Triple Threat)

What's Changed

🏗️ Model Improvements

• Evolved Super Resolution Neural Network from a Convolutional backbone to a Generative Adversarial Network backbone (#28)

🗃️ Dataset Changes

• Test csv download a bit better (3989fd7)

📦 Software Revisions

• Added unit tests to Continuous Integration workflow via a jupyter notebook (#10)
• Enabled conda activate and added cuda drivers (848b0a9)
• Updated docker buildpack-deps:bionic-scm base image version (#37)
• Added release drafter probot (103ff81)

Full list of commits made since last release viewable at compare view.

v0.2.0 (Two Is Better Than One)

What's Changed

🏗️ Model Improvements

• Peak Signal Noise Ratio (PSNR) metric added
• Model results can be compared with bicubic version of BEDMAP2
• FSRCNN was retrained many times on each new added dataset
• Built generator component of Super Resolution Generative Adversarial Network (SRGAN)

🗃️ Dataset Changes

• Use Inverse Distance Weighted Interpolation instead of Mean Elevation
• Change from 125m to 250m spatial resolution grid
• Added Pine Island Glacier dataset (#7)
• Added 200m REMA surface DEM mosaic (#8)
• Added 450m resolution MEaSUREs Ice Flow Velocity v2 geotiff (#20) and resampled to 500m
• Added CReSIS RDS data with fast vectorized tiling! (#21)

📦 Software Revisions

• Resolve various dependency issues by pinning versions
• Initial stab at Continuous Integration (#10 without tests)
• Tensorflow GPU added
• Trimmed docker image size
• Bump pipenv to 2018.10.13 (#31)

Full list of commits made since last release viewable at compare view.

v0.1.0 (Nimble One)

What's Changed

🏗️ Model Improvements

• Created Fast Super-Resolution Convolutional Neural Network (FSRCNN)

🗃️ Dataset Changes

• Used Rutford Ice Stream as starter sample dataset

📦 Software Revisions

• Setup conda and pipenv virtual environments
• Added various data science and geospatial dependencies

Full list of commits made since last release viewable at compare view.