senofsky / kestrel Goto Github PK

@kareemsasa3 is able to use an Ubuntu environment so there isn't any need for a Docker container.

Undo the merge.

We need to determine what video format should initially be supported.

The video format should be selected while considering:

• The video will be analyzed with software tools
• Video-handling software tools need to work with compression. Also see Machine Learning & Video Compression.
• The ultimate goal is performing object detection in real-time using a video camera
• That I am not knowledgeable on analyzing videos with software, so it is likely that other considerations need to be addressed.

The solution to this should be an informal research report.

If we want to detect litter, we need to have samples we can test against.

I arbitrarily picked 10; I think that is a "good enough" sample size for now. I am open to suggestions.

Videos should be in the format established in #1.

Create a library and add the first function. This will be the basis on which we develop the rest of kestrel.

I think this should just be a mock implementation of kestrel.trash_is_in_image().

We need to isolate and declare dependencies. @kareemsasa3 has already encountered issues trying to import cv2 on a Windows host.

We need to have a CI pipeline to maintain code quality and prevent regressions.

We need to add user documentation for kestrel.

We need to add black so we have a consistent coding style.

We need to add mypy for static type checking.

This will allow us to type check the code base and make it safer.

We need to add setup.py in order to create an official python package.

Add the following to the Makefile:

• prep-release
• upload-test-release
• upload-release
• clean