weather4cast2021_Stage1
3rd place solution for the Weather4cast 2021 Stage 1 Challenge
Dependencies
The code can be executed from a fresh environment using the provided list of requirements: conda env create -f environment.yml.
Inference
A script has been create to made predictions using a trained model on new data as per requirements detailed in competition: (#https://github.com/iarai/weather4cast#code-and-abstract-submission)
The model weights of the final submission for both core and transfer learning can be downloaded from https://drive.google.com/file/d/1mX4HMbf4QAW12DgAq1Bge1WT_5nAnz7N/view?usp=sharing
To run predictions on a test dataset use ('inference.py'). This should fine on a CPU machine
examples of usage:
- inference for Region R1
R=R1
INPUT_PATH=data
WEIGHTS=weights/Lomb_14.pth
OUT_PATH=.
python weather4cast/inference.py -d $INPUT_PATH -r $R -w $WEIGHTS -o $OUT_PATH -g 'cuda'
Train/evaluate a UNet
To replicate the training for the 3rd place solutions we recommend use of the training notebook (Training.ipynb).
Alternatively, a single script has also been provided (train.py).
In the actual competitions, the authors actually included an extra step to preprocess all the data and save them all on disk. This reduced the training time.
276 Dec 20, 2022
209 Jan 1, 2023
12 Oct 25, 2022
2 Jan 17, 2022
5 Nov 9, 2022
71 Dec 14, 2022
87 Dec 16, 2022
33 Dec 15, 2022
291 Dec 24, 2022
24 Nov 23, 2022
91 Dec 21, 2022
4 Jul 1, 2022
23 Mar 9, 2022
46 Dec 2, 2022
37 Oct 30, 2022
30 Nov 22, 2022
13 Jan 3, 2022
11 Mar 19, 2022
22 Nov 30, 2022