PFDet: 2nd Place Solution to Open Images Challenge 2018 Object Detection Track

  • 2018-09-04 02:59:07
  • Takuya Akiba, Tommi Kerola, Yusuke Niitani, Toru Ogawa, Shotaro Sano, Shuji Suzuki
  • 17

Abstract

We present a large-scale object detection system by team PFDet. Our systemenables training with huge datasets using 512 GPUs, handles sparsely verifiedclasses, and massive class imbalance. Using our method, we achieved 2nd placein the Google AI Open Images Object Detection Track 2018 on Kaggle.

 

Introduction (beta)

None

 

Conclusion (beta)

None