What question did this study set out to answer?

This research aims to enhance pedestrian retrieval accuracy through improved representation learning using SVD.

October 1, 2017

SVDNet for Pedestrian Retrieval

View Full Paper

Ask AI

Bookmark

Authors

Randomized trial demonstrates improved pedestrian retrieval accuracy, indicating effective deep representation learning strategies.

This research aims to enhance pedestrian retrieval accuracy through improved representation learning using SVD.
Proposes SVDNet leveraging singular vector decomposition to optimize fully connected layer training.
Implements restraint and relaxation iteration training scheme to ensure orthogonality in CNN.
Evaluates performance on Market-1501, CUHK03, and DukeMTMC-reID datasets.
Achieves rank-1 accuracy improvement from 55.3% to 80.5% for CaffeNet on Market-1501 dataset.
Increases rank-1 accuracy from 73.8% to 82.3% for ResNet-50 value.
Demonstrates reduced correlation among projection vectors leading to more discriminative features.

Sun et al. (2017) studied this question.

View Full Paper

Ask AI

Bookmark