Question about sodft_visma

[ghost]

Hi,
Thanks for sharing the code.

I have downloaded the VISMA dataset and successfully compiled both the dataset tools and this code (using build.sh). I also modified the paths in .json files, but I am not sure how can I use the apps.

I tried to executive the sodft_visma, but I found the code was stuck in socket->send(msg);. I assumed the msg will be sent to the faster RCNN model, and request the bbox. But I do not know how it works.

Shall I somehow build a Faster RCNN agent using the code in scripts, or I missed some points during the compilation?

Thank you in advance.

[feixh]

Yes, the tracker waits for bounding box messages from an external source. But you can modify the code so that the tracker operates by itself.

[ghost]

Thanks for your reply.

I looked into the code closely again. I found, actually in MOT_visma.cpp, you have already implemented a demo that directly read the pre-processed bbox and edge map without communication with Faster-RCNN. Is this the app used in the paper experiment?

I am trying to run this app, but now the code stuck in

VISMA-tracker/tracker/tracker_sir.cpp

Line 56 in b52c94b

ComputeLikelihood(level);

with level==1.
I am not familiar with the particle filter and do not know how to fix the problem.
Could you help me out?

BTW, there is a typo in

VISMA-tracker/tracker/tracker_utils.cpp

Line 25 in b52c94b

std::string full_path = StrFormat("%d/%d", root, cat_json);

which I believe should be StrFormat("%s/%s", root, cat_json);

[Hao-HUST]

Hi, @yangfengt123.
I met the same problem as you described. The mot bin stuck in

VISMA-tracker/tracker/tracker_sir.cpp

Line 56 in b52c94b

ComputeLikelihood(level);

.
More specifically, the code stuck in

VISMA-tracker/tracker/tracker_sir.cpp

Line 233 in 4116ba1

if (port_->handle() == 0) {

.
@yangfengt123 Did you solved the problem? @feixh Do you have any suggestions about that?