how to run the program

[yangailin]

Dear @alejocb,

I am trying to run the demo, but there has a problem: Segmentation fault (core dumped). Can you tell me how to slove it ? * DPPTAM is working *

* Launch the example sequences or use your own sequence / live camera and update the file 'data.yml' with the corresponding camera_path and calibration parameters * frames_processed -> 8.33333 % Segmentation fault (core dumped)

Thanks in advance~ Ailin

[alejocb]

Hi @yangailin ,

Did you have this segmentation fault with any of the provided sequences? Which sequence did you use?

Best, Alejo.

[yangailin]

For the provided two bag files, I have the same problem "Segmentation fault (core dumped)."

在 2016-05-06 22:23:04，"Alejo Concha" notifications@github.com 写道：

Hi @yangailin ,

Did you have this segmentation fault with any of the provided sequences? Which sequence did you use?

Best, Alejo.

— You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub

[alejocb]

Maybe the problem could be that the code only works with the openCV version of ROS. I have to fix this issue...

If it is the problem you could try this: https://github.com/alejocb/dpptam/issues/9

[sunghoon031]

Actually this might be a bigger issue than we think. Today I also ran the current master version on ROS, I got this segmentation fault. I ran the same code yesterday and it worked just fine. I re-cloned the master and it worked again.

I suspect that it's caused by the threading and copying of the pointer and accessing it while the source is deleted... I'm not too sure so I need to investigate it. Maybe the use of shared pointer could be the solution? I have a sequential implementation of DPPTAM (without any threading) and it's working fine without such issues.

[sunghoon031]

@yangailin For the time being, try to completely remove the ros DPPTAM workspace and reclone the project again. If changing the openCV version didn't work, this might solve the segmentation fault issue (temporarily)

[Sanghoon94]

Any solution for this problem? I haven't experience this problem when running on my desktop, but I am now trying it on NVIDIA TX1, and showing Segementation Fault same as Yangailin. Tried multiple times removing and cloning the project. Any suggestions? Thanks

[Sanghoon94]

I am trying on debugging the segmentation fault issue. I figured out that segmentation fault issue shows up at 299th line of SemiDenseTracking.cpp file which is "cv::Mat image_frame_aux = (*semidense_tracker->image_frame).clone();". So I've changed this into element-wise copy using

for(int i = 0; i < (_semidense_tracker->image_frame).rows; i++) for(int j = 0; j < (_semidense_tracker->image_frame).cols; j++) image_frame_aux.at(i, j) = (*semidense_tracker->image_frame).at(i, j);

And also the element-wise copy show segmentation fault, too. But the odd thing is it doesn't show error at the first index of the loop (i.e i=0, j=0). It kills itself with segmentation fault at random index of the loop. The index of error is different for every execution. So I am assuming somethings going wrong withe the shared access on the instance "semidense_tracker->image_frame" from different threads. I am trying to solve it using boost's mutex, but it's not working quite well. If anyone got any suggestions, I'd be really thankful.

Thank you.

[Sanghoon94]

Solved the problem.

During "cv::Mat image_frame_aux = (semidense_tracker->image_frame).clone();" is being processed by the semidense tracker, main thread's vo_system::imgcb function changes the content in semidense_tracker->image_frame .

So while semidense_tracker thread is reading the memory, main thread changes the same memory. That is where the segmentation fault was coming from. By locking the image fetching of imgcb when cloning the image in semidense_tracker, I was able to solve the problem.

[DunFenTiao]

i have the same problem of "Segmentation fault (core dumped)" with a live camera as input. what can i try to fix the problem?

[yangailin]

Dear Sanghoon94, Can you tell me how to locking the image fetching of imgcb when cloning the image in semidense_tracker?

[Sanghoon94]

I can't find the code I've wrote before, but I can briefly explain how I've solved it. The conflicting part was which caused seg fault in my case was:

1. "cv::Mat image_frame_aux = (*semidense_tracker->image_frame).clone(); " in SemidenseTracker
2. "vo_system::imgcb" in the main thread

So what you have to do is to skip or halt 1 while 2 is on process, and also skip or halt 2 while 1 is on process. I've tried using mutex locking in boost library, but it wasn't working well. So, I've naively implemented blocking using simple 'if' statements.

I think this is architecture (intel / arm) and machine-by-machine dependent problem, so I am quite not sure if this could solve your problem. Any way, good luck!

[DunFenTiao]

@Sanghoon94 i have the same problem on Segmentation fault (core dumped). thank you for your answer, but i am not familiar with multi-thread programming, how to implemented blocking using simple 'if' statements? would you explain more , provide any example that i can follow?^ ^ that will be a great help

[rnunziata]

I am also very interested in finding a solution to this I get fault. I tried a this solution and it did not work for me. On a upbeat note as this code seem to be unsupported is that ethzasl_ptam works .