For beta release i.e 0.2.0, tests have to be written.

To address issues #9 and #4, corrupted tiff files have to be identified by check-sums, re-triggering the download for this particular file.

At present, felicette waits for a RasterIOError, cleans the data directory, and triggers download for all the bands.

This also deletes non-corrupted data. Even if check-sums weren't used, an except statement should exist while processing bands 'individually' in felicette/sat_processor.py.

To address issue #7, the current solution is to preview image every time. An ideal solution would be to scan data for the band which is downloading first, or 'preview image' for inconsistencies. If there are some, try with a different product-id in the list of images it gets from sat-search.

Using PEP-518 to install numpy as build-dependency before running setup.py

Action Item: Right now, the options to search for a margin of cloud-cover, dates are hard-coded. These arguments have to be passed through CLI

Sentinel data is publicly available for download and usage by https://scihub.copernicus.eu/, and it gives RGB images at 10m resolution, while current data source Landsat-8 gives 15 at best with panchromatic resolution. The reason Landsat-8 was chosen over Sentinel is AWS instance of Landsat provides option to download each band separately. On the other hand, Sentinel-2 data has to be downloaded for all 13 bands, while we only require 3(R, G, B) for a bare minimum image generation with low resolution. (This requires a copernicus-scihub account)

Also partially free data is available at AWS for download(at a Requester Pays bucket).

Action Item: Both configurations need to be made available for usage with felicette

Status: Using copernicus-scihub account to download Sentinel data is not viable, because the downloadable file-sizes are ~1.3 GBs. Plus the user has to register in the portal. On the other hand, requester pays bucket is a more feasible option and already implemented with 0.1.2. This task item is henceforth marked 'Done'

There are other types of basic image-processing which helps classify urban areas, forest-fires, water-resources. An extensive list can be found here

Action Item: To write band-data processors for these categories of images.

Click has supports for sub-commands. At the current state of felicette, it uses only "options" to take user-inputs, which isn't good by design.

Action Item: Figure out and separate commands and arguments with felicette.