Progress/Experience using Chordata

daylanKifky

HalloweenBob the disconnecting issue which I can't seem to get past no matter what.

Do you mean the addon stopping because of the division by zero bug? Have you tried the patched version i described above?

HalloweenBob

Just tried the patched version of the addon.

SUCCESS!!!!!

It has been connected and been showing good smooth movement now for 10 minutes with no glitches or disconnects.

That's it! Now, I need to put together the whole chain and start moving around the model, then get to work in converting the output to a usable format for the software I use for the skulls.

Thank you daylanKifky and everyone else who contributed here.

Whatever was in that patch fixed this up good. Do we know why some people need the patch and others don't? Also, do we know if someone who wasn't having an issue used the patched version, would it give them problems, or just keep working. If it doesn't cause anyone a problem, perhaps the patched version should become the main version.

HalloweenBob

I was trying to put together the KCepters in the right order and assign them the proper jumpers for channel numbers, and came across this listing in the Default_biped .xml:

It says that the order for the arms is forarm, arm, then hand labeled 0, 1 and 2 or the first jumper, second jumper, third jumper, when it seems that the actual order should be arm, forearm, hand as shown in the diagram here: https://wiki.chordata.cc/wiki/File:DefBipedConf.png. If the forearm actually has a setting of 0, or jumper 1 as an identifier, that would mean that the KCeptor that comes first out of J1 should be identifier 1, then the next one in that chain should be identifier 0, followed by the hand at identifier 2.

What are the correct setting here?

I have to review all the instructions for moving the Default_biped model, because I thought that since my KCeptor model was reacting correctly to the motion of the KCepter, when I switched to the Biped model, the motion of my KCepter would move a portion of that model corresponding to J1, position 1 with identifier 0, which depending on where you look is either the left forearm or the left arm. I got no motion. Where can I find the instructions for creating motion with the default_biped model. Does it require a different command to the Pi? Do all the KCepters need to be connected? I think it's all working now, but I am just missing something. Also, another question. If I were only interested in capturing the motion of the head, would I only need a single KCeptor pluged into J2 and set to identifier 2 (the third jumper), or do I also need the Dorsal KCepter connected in order for the head to know where it is?

daylanKifky

Hi
There was a naming confusion from the beginning on our side. The Picture you linked has the first naming convention we used. See here for more details about it: https://forum.chordata.cc/d/98-build-log/19

As I said in the linked thread the naming conventions will change in the future, but the current armature works with the xml you pasted

In order to capture the complete body you should start the notochord with the default_biped.xml like this:

bin/notochord -c default_biped.xml <YOUR IP>

HalloweenBob If I were only interested in capturing the motion of the head, would I only need a single KCeptor pluged into J2 and set to identifier 2 (the third jumper), or do I also need the Dorsal KCepter connected in order for the head to know where it is?

If you want to get spatial translation of the head in reference to the hips you would need to use the three central KCs. If you just want the rotation of the head you can omit the other ones

HalloweenBob

OK, I tried to use the default_biped.xml tonight and couldn't get it to move at all. I thought I would start by trying just a single KCepter, the head. I plugged it into J2 on the hub and set the identifier to the 3rd jumper position (The closest one to the input socket) That seems to be where the xml expects to find it.

I used the following instructions to set everything up and run it:

CAPTURE PROCEDURE

Open New Project in Blender
Delete EVERYTHING. That box, the camera, the light, everything.
Select Window > Chordata > Add a Mocap Avatar to the scene
Select the "Mocap" tab that appears
There will be three nodes, the Notochord node, the Armature Node, and a Status Node
Add a new "Record" Node.
Connect the "Data Out" output of the Armature Node to the In on the Record Node.
Set the Target Object in the Record Node to the Chordata Biped.
Make sure the Notochord node is not in "Demo Mode"
Start Notochord on the Pi with ./notochord -c ../default_biped.xml <IP ADDRESS> Where "<IP ADDRESS>" is the actual IP address of the computer running Blender. No "<" symbols, just the number.
Hit "Connect" on the Notochord Node in Blender
Click "Show Advanced" on the armature Node to reveal Timed Calibration.
Click that button, making sure to set the timer for the appropriate number of seconds.
Look towards the SOUTH and assume T-Pose for Calibration.
When Calibrated, Hit "Record" on the Record node. And click it again to turn off the capture.
Each pass is saved as an "action." Select the "Active Action" you want from the pulldown menu in the node.
Hit "Play" on the Record Node to preview.
Select Chordata_biped in the Collection window of Blender. The motion capture data will show in the timeline.
Select "Export-> BVH" to make a BVH file to apply in another program.

In the SSH window, as soon as I sent the command, it started giving me error messages saying that it couldn't find all the KCepters that I wasn't using. I thought I would be able to only use one if I wanted to. It would seem that is incorrect. I did the timed calibration as outlined in the list above, and then started moving the KCepter, but it had no effect on the model.

Do I have to plug in every KCepter just to test the motion for one?

Perhaps I need to make my own xml that only includes the one KCepter, if I want to do that. I'm not sure. Ultimately, I want an xml that only contains the head, and the arms. For my application, those are the only moving points. The rest of the body remains static. I don't even need the hands. Just 5 KCepters (Although I may include hands for future expansion).

How should I be proceeding in order to make an arrangement like that to work?

daylanKifky

HalloweenBob Do I have to plug in every KCepter just to test the motion for one?

If you pass an xml file with the flag -c the notochord will try to find every KC described there, and assign the corresponding name to it. In order to capture just the head you can:

Modify the xml to include just that KC, in this case it will have a meaningful name (for example head)
Use the --scan function, in this case the notochord will look for connected KCs and assign generic names based on the location and value of the KC (for example kc_CH1-2)

When the data is sent to blender the armature will be moved if it receives a message from the notochord matching the name of one of the bones. The current armature does have a bone named head so you will see it moving if you used the first strategy.
You could also rename the bones on the armature to match the generic names given by the notochord to each sensor if you prefer the --scan method.

Let me know if this solves your doubts

daylanKifky

PS: The Dump node can be useful to better understand the messages you are receiving from the notochord. Take a look at this video we recently made where we explain how to add and manages nodes. together with some other basic blender actions.
https://forum.chordata.cc/d/136-new-video-blender-basics

HalloweenBob

I know it has been talked about here on the forum, but can you point me to a thread or something on the Wiki that tells me exactly what the required parts are, and the correct syntax to a custom xml file so I can make one using just the parts I will need? Or even a link to the default xml that I can just edit out the parts I don't need. If I remove all the parts related to the legs for example, do I remove all references to them or just the KCepter branches leaving the main Branch names there, but empty? I don't know if it looks for the entire structure of the xml to work with the Biped model or if it doesn't need it all. I hope I'm explaining this correctly.

HalloweenBob

I found a copy of the default_biped.xml file here on the forum. I will modify it for what I need and post it here. Hopefully someone will be able to tell me if I have done it correctly, or broken it somehow. I won't be able to test it until late tonight, so the feedback will help me save trial and error time tonight.

HalloweenBob

OK, here's what I have. I only care about the arms and the head. I don't really care about the hands, but have left them in just in case I add motion to those in the future. As a reminder, I am using this to program lifesize skeleton animatronic props. They are static except for the arms and the heads. The bodies are clamped to stands, so there no way for them to move their legs, torso, the dorsal area, or anything else other than the two arms and 3 axis motion of the head (and their jaw, but that is being handled in a different way).

I will post what I did to modify the xml file, and would ask that someone verify that it would indeed work, or if I removed too much of the main structure and would cause it to crash.

I also have questions about the IP address embedded in the xml file. Is that supposed to be an address in my local system? Is is supposed to be the address of the device running blender, or the Pi? I ask because it looks like a local network address, but it is certainly not my local network address.

My final question is do I name this file something unique, and save it to the notochord/bin directory? I don't know the SSH command to move or copy files, so I would need that. Once in the correct directory, I assume I would follow those numbered directions in my previous post from yesterday and use the -c command just modifying it to reflect the new filename I create. Does all that sound right?

Here is the modified xml file I created

modified-biped.txt

2kB

This is the text from it, in case it didn't show up in a readable format:

<!-- Chordata.xml -->
<Chordata version="0.1.0"> 
    <Configuration>
            <KC_revision> 2 </KC_revision>
            <Communication>
                <Adapter>
                    /dev/i2c-1
                </Adapter>
                <Ip>
                 192.168.85.123
                </Ip>
                <Port>
                    6565
                </Port>
                <Log>
                    stdout,file
                </Log>
                <Transmit>
                    osc
                </Transmit>
                <Send_rate>
                    50
                </Send_rate>    
                <Verbosity>
                    0
                </Verbosity>
            </Communication>
            <Osc>
                <Base>
                /Chordata
            </Base>
                
            </Osc>
            <Fusion>
                <Beta_start>
                1.0
            </Beta_start>
                <Beta_final>
                0.2
            </Beta_final>
                <Time>
                5000
            </Time>
            </Fusion>
        </Configuration>
    <Armature>
            <Mux Name="main" id="0">
                0x73
                <Branch Name="left_arm" id="1">
                        CH_1
                        <K_Ceptor Name="l-forarm" id="2">
                            0
                        <K_Ceptor Name="l-arm" id="2">
                                1
                                <K_Ceptor Name="l-hand" id="2">
                                    2
                            </K_Ceptor>
                        </K_Ceptor>
                    </K_Ceptor>
                </Branch>

                <Branch Name="body" id="1">
                        CH_2
                    <K_Ceptor Name="head" id="2">
                                2
                    </K_Ceptor>
                </Branch>

                <Branch Name="right_arm" id="1">
                    CH_3
                    <K_Ceptor Name="r-forarm" id="2">
                        0
                            <K_Ceptor Name="r-arm" id="2">
                            1
                                <K_Ceptor Name="r-hand" id="2">
                                2
                                </K_Ceptor>
                            </K_Ceptor>
                    </K_Ceptor>
                </Branch>
            </Mux>
        </Armature>
</Chordata>

Thank you!

daylanKifky

HalloweenBob I will post what I did to modify the xml file, and would ask that someone verify that it would indeed work, or if I removed too much of the main structure and would cause it to crash.

It's looking right to me. It will try to read 7 sensor in total, the complete hands and the head (as you described). By the way: at the beginning of the execution the notochord will validate the structure and syntax of the XML, so if there's an unperceivable syntax error it will inform you.

HalloweenBob I also have questions about the IP address embedded in the xml file. Is that supposed to be an address in my local system? Is is supposed to be the address of the device running blender, or the Pi? I ask because it looks like a local network address, but it is certainly not my local network address.

It is the same address you normally pass to the notochord as a command line option. We normally don't mention this because most people don't tweak the XML, but you can just write the address of the PC running blender there instead of passing it to the notochord every time. Many other options can be set in the XML as well as in cmd line but the latter ones (if present) will overwrite the former ones.

NakedRabbit

Is it me, or does this part look wrong?

               <Branch Name="body" id="1">
                        CH_2
                    <K_Ceptor Name="head" id="2">
                                2
                    </K_Ceptor>
                </Branch>

I ask because I had issues with this exact same sort of thing. This would indicate that J2 on the Hub has one Kceptor with the jumper at J5 set to the THIRD position. My understanding was that you should have it set to the FIRST position for the first Kceptor on the line, and then the XML should have a value of "0" for that Kceptor. But maybe it does not matter as long as it is consistent? I dunno...

HalloweenBob

NakedRabbit That's a good question. I assumed that the jumper identifies that KCepter as the "Head" so it knows what thing to move around, but maybe it's just the name or the ID line.

By the way, that particular branch only has two items in it in the regular xml. They are set to the 2nd and 3rd jumper. That branch does not have a KC set to the 1st jumper.

If I have an issue with it, I'll try changing the jumper, but if that were the case, it wouldn't work in the default_biped.xml file either.

I'll know later tonight when I get home and get to try it out.

HalloweenBob

OK, I've learned a lot and I'm almost there.

I made a new xml file called head_biped.xml which only has the head KC attached. Then I had to find a way to move it to the Pi and put it in the right directory. A little research on Google and I found a program called WinSCP. It was on a website called ninite.com. It's basically an FTP program showing you all the folders and files in your local computer on one side and all the folders and files on the Pi on the other. You can just drag and drop files from one side to the other to move files around. I'm just posting this so others who might have similar questions can save some time and do what I did.

I set up blender, and ran the -c command using the new name for the xml and then connected in blender and it worked to some degree. I feel I am really close right now. The head started moving with the KC. It looked kind of like Linda Blair in the Exorcist, but it was moving.

I did the timed calibration, but things are still out of wack.

I downloaded a compass app so I could face SOUTH while running the calibration, but I was using my desktop computer with a monitor mounted on the wall that faces pretty much North, so it was hard to calibrate while facing the opposite direction as the screen.

I'm not sure how I should orient the KC on my head. When I attach it so that the model is looking straight forward, I can rotate my head back and forth, and it seems to move correctly, but when I nod my head, it tilts instead, and when I tilt it nods, but there is no way to orient it so that those motions are correct without affecting the rotation.

The axes also don't seem to relate to any normal position of the KC. To get it to nod straight up and down, it has to be at a 45° angle.

So I still don't know how to calibrate it correctly I guess or the right way to orient it on my head. Should it be on my forehead with the in and out connectors facing away from my head on the top? Should it be on top of my head like a hat? I'm not sure how it should be attached.

If I can just get the calibration correct and have it move the way my head is moving, I'll be done.

The other question I have is once calibrated, do I always have to only face south to use this? I understand that since I don't have the dorsal or base connected it has to assume that I am always facing South. The body won't move, but the head will spinning it around backwards. I'm good with that, because my skeletons can't move around freely anyway, so I am happy scoping out which way is south and setting up in front of a laptop, or maybe I just connect the rest of the KCepters and get full motion, but I think that may confuse my conversion process to my servo controllers.

So I need to figure out the correct orientation of the KC on my head and why it is moving different axes than I am after calibration. Once I get this working, I'll add in the arms and see how that goes.

Note to NakedRabbit, The 'body' branch that you questioned is correct. Although it's not moving quite the way it should be. it is moving the head using the code exactly as you pasted in a couple posts up.

HalloweenBob

Does anyone have any hints on the right way to orient the KC on my head, or why it seems to nod when I tilt and tilt when I nod? I'll be trying it all again tonight, using a compass to face south. Maybe that will make a difference. Please feel free to share whatever you had to do in order to get the head motion correct. Thank you.

daylanKifky

HalloweenBob
The orientation is not important. It looks like you are facing east or west when calibrating. Were you roughly facing south?

HalloweenBob

Yes, I was using a compass so I would know I was facing south. I did the calibration again last night, and it seemed to work for a moment or two. Then I moved, and when I went back to facing the same way, it was no longer facing front and it seems like all the positions had drifted. Nodding my head showed on the screen as more of a diagonal or combination between tilting and nodding and I had to keep moving which was I was facing to keep the model looking straight forward.

daylanKifky

Is the head bone green when you do the pose calibration?
otherwise try rotating the the sensors around at the beginning of the capture as I do at 7:20 in this video

HalloweenBob

I watched the video, but am still having trouble. Here is a screen capture of Blender with my narration of what I was doing.

I am still trying to get this to work with just one KC. (The Head) No matter what I did, I could not get the line in the head to turn green. During some attempts when I was wildly shaking my head it would go green for a second or two, then go back to red.

I was facing South. I did stand still and even raised my arms for some reason even though I only had one KC connected.

This is how I had my KC attached to my head.

Sorry for the huge image. I bought an LED head lamp and removed the light and just used the strap through the 3D printed KC case.

I wonder if I didn't calibrate the KC correctly initially. I have watched the calibration video many times, but I'm still not 100% sure of the exact procedure.

I'm not sure if I should click to start the process while it's still sitting on the desktop, then pick it up and start turning it, or do I click to start the process of turning it after I am holding it and start turning right away. It seems to suggest that I start with it lying flat on a table (After the initial click where you just leave it on the table and it gets it's baseline settings). While it's lying there, I click to start the calibration . After it starts, I pick it up and rotate it clockwise (I'm not sure if the direction makes a difference) until it has been turned 360°, then flip it on it's edge (Not sure if I flip it so that J1 and J2 are facing towards me or away from me or if it matters) and rotate it in that orientation 360°. Then, I'm not sure if I should set it back down on the table first and then click to stop the calibration, or just click to stop the calibration as soon as I'm done rotating it. I'm also not sure if the rotations are supposed to be 1 smooth uninterrupted motion, or if I turn it 1/4 of the way, then stop for a second, then to 1/2 way and stop, then 3/4 and stop then 100% and stop.

I have lots of questions about the calibration, and since there's no message telling you if it was done correctly, I really don't know. I do know that something isn't right, and if I can't get even one single KC to react properly, I know I'm going to have trouble hooking up a whole set.

Any ideas?

daylanKifky

Hello!
Thanks for the video, it's always easier when I see how things are working.

The red bone indicates that the sensor fusion is getting a big error on the estimation, we should find out why.
Regarding calibration: the speed or rhythm of the rotations doesn't matter. The important part is that the sensor should "look" at a given point in space from all possible angles. From how you are describing it seems like you are missing a few flips on the edges. It is a good idea to do at least 6 turns of 360º with the sensor orientated like:

components side facing up
components side facing down
right edge facing up
left edge facing up
top edge facing up
bottom edge facing up

After doing an horizontal 360 turn with the KC starting in these positions you can conclude the calibration.

There's no message explicitly telling you it was correct, just that the calibration was recorded to the EEPROM. This is vaguely inspired on the way UNIX commands work, many of them only print messages when there's an error. Not a premeditated choice, one just get used to that way of thinking and programming.

But!

Even with the a non completely proper calibration the KC should be able to live-calibrate itself when you move it around. There might be something preventing it from getting proper measurements. For example I see a screw in the bottom of the case you are wearing in the head. Try calibrating and testing the sensor without metal pieces in the proximities.

If you still get a red bone please attach a Dump node to the notochord node, enable only the extras stream and paste what you get in the blender console (should be a bunch of messages with the /Chordata/extra/<bone>/amcov datatarget)
(note: it's a good idea to use pastebin.com or similar to paste long dumps. Just "Create a new paste" and include the link to it on your post)

« Previous Page Next Page »