Hi all, we just wanted to highlight that the Future Work section of the documentation has received a significant update.
If you navigate to the various sub-sections (e.g. Detect Local and Global Flow), you will find descriptions of the key projects we want to work on next - or that would be great contributions if anyone out there wants to work on them. Note that many of these topics are substantive, and could be written up as a paper or master’s thesis.
Please feel free to ask any questions you may have about the sections, we can always provide more details if it is helpful.
Re. actually contributing, we would recommend that you open an RFC to describe your proposed implementation if you would like to work on one of these areas. You can read more about RFCs and contributing here.
"… we still need to implement some more pre-requisite features for Monty to be useful in a wider range of applications (like being better at dealing with multi-object environments, modeling compositional objects, and using policies to manipulate the world) …
This makes me wonder about setting up a way to refer to
specific “pre-requisite features” that would be needed for a given sub-project. Suggestions, anyone?
that’s a good point. The thing that comes to mind is this overview sheet: Monty Project Overview - Google Sheets We link to it in our “Future Work” section of the documentation. There we also write out details on each of the items in the table.
If you have an idea for a project, you could look in the “Capabilities” column to see how far we are on the capabilities required for this project. Each capability has a #label associated with it and we tag the different items in the table on the left accordingly. That way you can see which specific improvements we are planning on for the different capabilities and how you could contribute to improve this capability. Of course, there are many things that are not in the table. We only add items into the left side of the table once we have a concrete idea of how they would be implemented. For some capabilities (like modeling object behaviors, transformations, and scale) we are still in the brainstorming/research/idea generation phase.
If you scroll down to the milestones below the table you can see which capabilities our internal team plans to tackle next and how we plan to do this. However, we welcome external contributions on all of these and if you start work on one of these items (after successful RFC merge) we will even add a player icon for you to this sheet.
I think the items in that spreadsheet column will work nicely. I’d like to use hashtags to keep the entries unambiguous, then add descriptive text for clarity. In terms of the gardening robot, I might put:
Required Monty Capabilities
#compositional - Plants will have various components.
#multiobj - Various types of plants will appear in the same image(s).
#realworld - The robot will need to move about, aim its camera(s), and perhaps employ a vinegar sprayer.
