CaptainEven/MCMOT

Bad demo results!

Opened this issue · 2 comments

Hello,
Thank you for the great repo, the Gifs are very cool,
But when I tried to run it with those 2 models (mcmot_last_track_hrnet_18_deconv.pth and mcmot_last_track_resdcn_18_visdrone.pth)

the results were so bad, with the first one:
image

image

and with the drone model the bbx did not fit on the pedestrian and also flicker a lot !
image

Is this the normal case or am I missing something,

This is how I ran the model:
python3 demo.py
--id_weight 1
--load_model ../models/mcmot_last_track_resdcn_18_visdrone.pth
--input-video ../videos/sample.mp4

Do I miss any other arguments ?

Did you solve this problem? I have similar issue and tried many things found in issues, but still my MCMOT put boxes everywhere but not over objects.
image

+1, same problem, not detection al all.

00018

here is my cmd:

python demo.py  --load_model mcmot_last_track_resdcn_18_visdrone.pth  --input-video /media/sdb/long/deta/test_videos/two_man_climbing.mkv