Reproduce “Multispectral Deep Neural Networks for Pedestrian Detection” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Jingjing Liu, Shaoting Zhang, Shu Wang, Dimitris N. Metaxas
Multispectral deep neural networks were developed to improve pedestrian detection by leveraging complementary color and thermal imagery. Fusing middle-level convolutional features (Halfway Fusion) achieved the lowest miss rate of 36.99% on the KAIST dataset, outperforming single-modality detectors by about 11%.
Key insights
- Color and thermal images exhibit strong complementarity for pedestrian detection, with each modality excelling under different lighting conditions.
- Fusing these multispectral inputs within deep neural networks significantly improves detection performance over single-modality approaches.
- The specific stage of fusion within the deep neural network hierarchy profoundly impacts performance, with middle-level feature fusion providing the best synergy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.