CaptainEven / MCMOT

Real time one-stage multi-class & multi-object tracking based on anchor-free detection and ReID
MIT License
383 stars 82 forks source link

Can't find clear instructions on how to run the demo. #48

Closed joaqo closed 3 years ago

joaqo commented 3 years ago

Hi, I am trying to run the demo on a traffic video, but all my tests have returned videos with no detections, or videos with random detections. The environment I working on works perfectly with FairMOT, but doesn't work on MCMOT.

The command I am currently using is:

python demo.py --load_model ../models/mcmot_last_track_hrnet_18_deconv.pth --conf_thres 0.4 --input-video /path/to/video.mp4

The video processes with no errors and it also loads the checkpoint just fine, but it does not find any cars on several videos I've tried.

Could you post step by step instruction on how to run the demo on a video?

Thanks!

CaptainEven commented 3 years ago

@joaqo I'm sorry for the messy structure of the repo, and it will be improved in the future if i have more free time. Following the readme, you will run the demo successfully. Just like:

python(python3) ./demo.py

after you configure the opts.py. Wish good luck!

joaqo commented 3 years ago

I ended up making it run with the following command for anyone interested:

python demo.py --load_model ../models/mcmot_last_track_resdcn_18.pth --input-video /path/to/video.mp4

It seems the model was the problem, mcmot_last_track_hrnet_18_deconv.pth doesn't work and mcmot_last_track_resdcn_18.pth does. Do you know why this could be @CaptainEven ?

austinmw commented 3 years ago

@joaqo did you try switching arch? --arch hrnet_18

joaqo commented 3 years ago

@austinmw That was it! Now it's working fine, thanks a lot. For anyone reading this issue, make sure to match your --arch argument with the checkpoint you are using.