-
Please ask the author if he has verified the effectiveness of the Unipad framework on other backbones.
For example resnet-50. If so, what can I do to verify that resnet-50 is pre-trained under uni…
-
During the benchmarking meeting today, a number of topics was discussed regarding benchmarking. This issue summarizes/expands these discussions. This documents will be updated after meetings:
Three p…
-
### Prize category
Best Plugin
### Overview
This project includes the creation and integration of three unique models inside the Impact Engine Framework (IF) to support green computing practi…
-
[Quentin Paletta et al.'s 2021 "ECLIPSE" paper](https://arxiv.org/abs/2104.12419) show that it's a good idea to predict cloud _masks_, because these are concise representations of the sky (with the ca…
-
PyRTM contains functional python wrappers for SMARTS and SBdart, and is capable of modelling spectra in clear-sky and cloudy conditions. This would be a great set of functionality to incorporate into …
-
Hi,
I have time data and split to train and test (keep it unseen) by slicing the `df` from the end part. I used your pipeline over `data_train` and tried to forecast as length as `data_test` unsu…
-
-
update by @fkiraly - summarizing the below, I think this is a combination of feature request - "forecasters should support forecasting on a subset of instances, if panel data" - and frustrated user ex…
-
Various ones to try include https://github.com/thuml/predrnn-pytorch and ConvLSTMs, and potentially GANs as shown in https://arxiv.org/abs/2104.00954
-
Forecasts in the EFI NEON challenge are all explicitly site-based timeseries.
With the widespread availability of remote sensing & imagery data and the importance of understanding spatially expli…