TMO-Det: Deep tone-mapping optimized with and for object detection

Kocdemir I. H., Koz A., AKYÜZ A. O., Chalmers A., ALATAN A. A., Kalkan S.

Pattern Recognition Letters, cilt.172, ss.230-236, 2023 (SCI-Expanded) identifier identifier

  • Yayın Türü: Makale / Tam Makale
  • Cilt numarası: 172
  • Basım Tarihi: 2023
  • Doi Numarası: 10.1016/j.patrec.2023.06.017
  • Dergi Adı: Pattern Recognition Letters
  • Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Academic Search Premier, Applied Science & Technology Source, Compendex, Computer & Applied Sciences, INSPEC, zbMATH
  • Sayfa Sayıları: ss.230-236
  • Anahtar Kelimeler: Generative adversarial networks, High dynamic range, Low dynamic range, Object detection, Tone-Mapping
  • Orta Doğu Teknik Üniversitesi Adresli: Evet

Özet

Detecting objects in challenging illumination conditions is critical for autonomous driving. Existing solutions detect objects with standard or tone-mapped Low Dynamic Range (LDR) images. In this paper, we propose a novel adversarial approach that jointly optimizes tone-mapping (mapping High Dynamic Range (HDR) to LDR) and object detection. We analyze different ways to combine the feedback from tone-mapping quality and object detection quality for training such an adversarial network. We show that our deep tone-mapping operator jointly trained with an object detector achieves the best tone-mapping quality as well as detection quality compared to alternative approaches.