CaptainEven / MCMOT

Real time one-stage multi-class & multi-object tracking based on anchor-free detection and ReID
MIT License
390 stars 83 forks source link

No detection or tracking bounding boxes in the output video #73

Open AhmedAAkl opened 3 years ago

AhmedAAkl commented 3 years ago

Hi @CaptainEven, Thanks for this nice work,

After training the model on the VisDrone dataset 10 classes with this command. I used this file gen_dataset_visdrone.py to generate the dataset format for training.

python train.py --task mot --num_epochs 60 --lr_step '50' --data_cfg '../src/lib/cfg/cars_data.json' --load_model '../models/all_dla34.pth' --output-root '../results' --arch dla_34 --batch-size 6

the training loss decreased properly by nearly 0.5146. but when I tested the model in a video, it was unable to detect any objects even when I used --conf_thres 0.1 Is there any missing steps?

any help will be appreciated.