missing param.prototxt about hityavie

[hitdshu]

Hi,
Im tryng to test with on the suggest dataset but the result is not good:

This is my configuration file:
https://github.com/ceccocats/hityavie/blob/master/proto/param.prototxt
can you please check it?

Also I often get this error with different number of features (ex 200):

WARNING: Logging before InitGoogleLogging() is written to STDERR
F0701 13:11:17.967087  3661 problem_impl.cc:671] Parameter block not found: 0x3567bd8. You must add the parameter block to the problem before you can set its local parameterization.
*** Check failure stack trace: ***
Aborted

Any hints?

Thank you very much

Thanks for your interest and work. It's my pleasure to check the error and relating settings. Since I don't have much spare time on workdays, I'll try to debug it on this weekend. Let's hope we can solve this problem.

Deshun

[ceccocats]

I achived better results

The first lap is almost good, but the sencond fail:

How I can debug this? Is it the Imu? is it the feature tracker? is it the optimization?
There are a lot of variables...

[hitdshu]

Hi Francesco,

Thanks for your interest. I'm afraid I've lost the "param.prototxt" since I use a new laptop now and some data in my old laptop may be missing now...

I use the UZH_FPV2019 data for prototyping this method. See https://github.com/uzh-rpg/IROS2019-FPV-VIO-Competition . And maybe you could reproduce the param.prototxt from that data(This should be a big challenge). And may I suggest you to use other vio systems like Open_VINS/BASALT/VINS-MONO,...etc.

Good luck,
Deshun

[ceccocats]

Thank you,
I like yours because is small and simple to understand, I already adapted your code to read my dataset but the prototxt is full of parameters that i have to try; maybe I will just use it as a tutorial.
The code does not have a license, is it free?

[hitdshu]

Thank you,
I like yours because is small and simple to understand, I already adapted your code to read my dataset but the prototxt is full of parameters that i have to try; maybe I will just use it as a tutorial.
The code does not have a license, is it free?

No license. Absolutely free. And you can use it or modify it as you like.

[ceccocats]

Hi,
Im tryng to test with on the suggest dataset but the result is not good:

This is my configuration file:
https://github.com/ceccocats/hityavie/blob/master/proto/param.prototxt
can you please check it?

Also I often get this error with different number of features (ex 200):

WARNING: Logging before InitGoogleLogging() is written to STDERR
F0701 13:11:17.967087  3661 problem_impl.cc:671] Parameter block not found: 0x3567bd8. You must add the parameter block to the problem before you can set its local parameterization.
*** Check failure stack trace: ***
Aborted

Any hints?

Thank you very much

[hitdshu]

It seems I can't download the dataset....

[ceccocats]

Try this:
https://cloud.hipert.unimore.it/s/j348bSpZHWNZ8XN

[hitdshu]

Thanks. I downloaded it and use it for the program. The result is terrible. I think it might be the wrong variance config of the imu related noise.

----Your imu noise setting
np {
acc_noise: 0.1;
gyr_noise: 0.05;
ba_noise: 0.5;
bg_noise: 0.5;
}

We might check this link for some information https://github.com/ethz-asl/kalibr/wiki/IMU-Noise-Model

Since it's late now, I have to stop debugging and go to bed. And I'll continue to debug the program tomorrow.

Deshun

[ceccocats]

Thank you!
I used this calibration file from the dataset providers for the IMU:
results-imucam-..indoor_forward_calib_davis_imu.txt
Downloadable from(end of page): http://rpg.ifi.uzh.ch/uzh-fpv.html

ba_noise and bg_noise is completely random I dont know what is their meaning

Also, from the calib file gravity should be "-9.30066998"

[hitdshu]

Well, that's impressive... Surely there are a lot of variables, and actually I've forgotten their true meaning too... I'm greatly surprised by your debugging result since this project seems like a stranger to me now.

Thanks for your great work. I devote recent weeks to my another project https://github.com/hitdshu/hitnlls, so I didn't spend much time debugging your problem..., sorry and hope you won't get disappointed.

I didn't take serious attention to hityaive while writing them since I thought perhaps no one would use them... Now, perhaps I'm wrong and at least someone(perhaps only you) is using them...

And I'll start to rewrite hityavie project in a week or two, after finishing some unit tests and examples for hitnlls project.

[hitdshu]

By the way, feel free to modify the code. Since it doesn't have a license, you can use it as your code or everything else.

[ceccocats]

Thank you very much, also for me this is a side project I'm working on this in the spare time. I'm also looking at openvins.

By the way, feel free to modify the code. Since it doesn't have a license, you can use it as your code or everything else.

Actually if it doesn't have a License all rights are yours by default, this is why I asked you some before. If you want to give free usage you should attach an open license like MIT.

[ceccocats]

It's funny because I'm also working at a solver. With the same interface of ceres-solver but completely on GPU

[hitdshu]

Thank you very much, also for me this is a side project I'm working on this in the spare time. I'm also looking at openvins.

By the way, feel free to modify the code. Since it doesn't have a license, you can use it as your code or everything else.

Actually if it doesn't have a License all rights are yours by default, this is why I asked you some before. If you want to give free usage you should attach an open license like MIT.

Ok. Now I've added a license https://github.com/hitdshu/hitnlls/blob/master/LICENSE, copyed from https://github.com/pjreddie/darknet/blob/master/LICENSE.fuck. Hahaha.

[hitdshu]

It's funny because I'm also working at a solver. With the same interface of ceres-solver but completely on GPU

That's an interesting project. Looking forward to your solver.