search

daip13 / LPC_MOT

This is the code for the paper "Learning a Proposal Classifier for Multiple Target tracking"
MIT License
74 stars 11 forks source link

ReID Model Inconsistency #9

Open ocetintas opened 2 years ago

ocetintas commented 2 years ago

I would like to calculate ReID features by myself by running ReID_feature_extraction.py However, this file requires the pretrained weights called model_final.pth as shown here.

The ReID models you have provided are called "with_mot17_data" and "wo_mot17_data"; however, you don't provide a "model_final.pth". Which model should I use to achieve the numbers you report in the paper?

Moreover, in the paper you report that you train your reid model on MOT17 and MOT20. However, none of the provided models seem to be trained on MOT20 according to the naming scheme. Can you please provide the model also trained on MOT20 as well?

daip13 commented 2 years ago

I would like to calculate ReID features by myself by running ReID_feature_extraction.py However, this file requires the pretrained weights called model_final.pth as shown here.

The ReID models you have provided are called "with_mot17_data" and "wo_mot17_data"; however, you don't provide a "model_final.pth". Which model should I use to achieve the numbers you report in the paper?

Moreover, in the paper you report that you train your reid model on MOT17 and MOT20. However, none of the provided models seem to be trained on MOT20 according to the naming scheme. Can you please provide the model also trained on MOT20 as well?

Hi, for MOT17 testing set, you can use the model in "with_mot17_data" to achieve the performance in the paper. The "model_final.pth" should be the model for MOT20. I will try my best to find it, but I can't guarantee 100% of it. It was too long ago. Maybe you can train the reid by yourself with the scripts in fast-reid.