Skip to content

Latest commit

 

History

History
20 lines (15 loc) · 839 Bytes

README.md

File metadata and controls

20 lines (15 loc) · 839 Bytes

competition_iqiyi

ref

Experiment Report

method train loss TRAIN mAP VAL mAP TRAIN Precision VAL Precision Test mAP
mtcnn + insightface + means knn (baseline) none none 0.771192 none 0.82217 none
mtcnn + insightface + our FC + means knn 2.01 none 0.365657 none 0.4204 none

InsightFace+FC Usage:

Train steps:

  1. Git clone the InsightFace_TF reposity.
  2. Train the model by train_fc_only.py script in InsightFace repo.
  3. Evaluate the validate set by eval_map_fc_only.py script.