search

IS2AI / trimodal_person_verification

This repository contains code and data for "On the Multimodal Person Verification Using Audio-Visual-Thermal Data"
MIT License
11 stars 0 forks source link

Recurrence problems #1

Open xyz-hcy opened 2 years ago

xyz-hcy commented 2 years ago

Dear authors: I use the same data and model to reproduce the following commands:

python trainSpeakerNet.py --model ResNetSE34Multi --modality rgb --log_input True --trainfunc angleproto --max_epoch 600 --batch_size 100 --nPerSpeaker 9 --max_frames 200 --eval_frames 200 --weight_decay 0.01 --seed 1 --save_path exps/rgb/exp1

The model training results are shown in the figure below. Loss decreases, TEER/TACC increases.The VEER of the validation set is stable at 27. What is the reason for the result？ image Thanks in advance.

Regards

songhaozhen commented 2 years ago

hello,could i add you VX or QQ,i think i lost the trained model and train list,thank you very much QQ2250050388