Create, review, and submit the video attachment

[spmaniato]

I submitted the paper last night: ICRA16_2184_MS.pdf You can also get it from PaperPlaza.

I now have until Sunday 09/20 to create and submit the video attachment (according to some pretty strict guidelines #6 regarding the size of the video file).

I'll put together a draft by tomorrow at noon and ping you for a quick round of comments.

[spmaniato]

On a second thought, I'd rather upload a partial draft now and get some early feedback. So here's a draft of the first demo: Experimental Demo 1 (no narration yet, sorry) I highly recommend that you download the file. The quality is pretty bad when you watch it via Dropbox's video player.

@hadaskg @dcconner @pschillinger, Looking forward to your comments :smiley: (I already know the fonts look horrendous btw :stuck_out_tongue_closed_eyes:) Here are some specific questions:

• Do you agree with my overlaying of some info (e.g., I, G, Out(Exec)) over the video?
  • I'm trying to relate what's going on to the paper and its notation.
• Does the video become too "busy" once all 3 views are up? (FlexBE, MainView, Camera) I think it's OK because once execution starts, they don't have to pay as much attention to FlexBE anymore.
• Should I speed up the part where the operator (aka, me) is aligning the template?
  • There is no length/duration restriction, so don't worry about that.
• Do you think the beginning (paper title and author names) is OK? I couldn't find any guidelines.

Thank you!

[pschillinger]

I like it.

• Yes, I think the overlay is good. Otherwise it would be much harder to see what you write and this is the essential part.
  • Maybe you can add what the symbols mean (e.g. "Initial conditions: I = {...}" instead of just "I = {...}"). This will help if a person watches the video before reading the paper and recalling symbols.
• The 3 views are fine, I think.
• I'm afraid speeding it up would look strange, especially when rotating the 3D view. On the other hand, right now template alignment takes around 30% of the time without being relevant itself. Not sure about this question.
• Beginning looks good if there are no guidelines.

Some other things I noticed:

• A person who doesn't know our workflow of manual template alignment might not know what is going on when you switch to the 3D view and click around. Maybe another text hint? (btw if there was sound, I would have watched the video initially with speakers muted because they always are, so don't just rely on what you say for general understanding)
• It's good you wait a short time at 0:45 before confirming the transition. This helps the viewer to keep up with what is happening.
• However, later some of the transitions could be autonomous, I would say. Otherwise most of the viewers will probably think "Oh, he has to click a lot, it's not that autonomous". Just my experience on their initial reaction when I explain people how the Autonomy Level concept works.
• I would rather speed up the re-arranging part of the state machine (or maybe even cut it out?). I also tested an auto-layout algorithm, but unfortunately, it doesn't look nice.

[spmaniato]

Thanks Philipp! I'll take most of your comments into account :smiley:

However, later some of the transitions could be autonomous, I would say.

Unfortunately I can't change things like that because this screencast was taken back in July. Since it's an experiment on Atlas, not in simulation, I can't repeat it.

[dcconner]

I think the overlays are helpful and would like to see more to describe what the operator is doing. (e.g. when you are looking at the single higher-level state and click to access the synthesized code, manually adjusting the template position.)

Narration will help , but having a few key phrases will help orient the viewer even without sound.

Looks good.

[spmaniato]

I've incorporated a couple of your suggestions and also added the 2nd demo: video v2 (full draft)

Could I get some more feedback please? Especially about the 2nd part, which demonstrates runtime modifications via synthesis (and viewers might have a harder time understanding).

Btw, I'm planning on doing the narration last, since it depends a lot on things like video length and speed. Also, after adding the 2nd demo, we are at 15MB (the limit is 10). I'll see what video editing settings and options I can mess with, but keep in mind that we may need to cut stuff. The template alignment marathon will be the first to get chopped :stuck_out_tongue: Other suggestions are welcomed. Thanks :smiley:

[spmaniato]

Latest full draft: video v4 Looking forward to any feedback. Now is the time to also comment on stylistic things, like my choice of fonts :smile:

It's about 30 seconds shorter than before because I skipped or sped up some parts (state machine rearrangement, template alignment, etc.) Current file size is 9.5MB

We are also considering not doing any narration at all. Assuming that the subtitles provide enough info.

Thanks!

[spmaniato]

This is the video that I ended up uploading to PaperPlaza: video v6