Deep Similarity Metric Learning for Real-Time Pedestrian Tracking
This work addresses pedestrian tracking for surveillance or autonomous systems, but it is incremental as it builds on existing tracking-by-detection methods with a learned appearance metric.
The paper tackled pedestrian tracking by integrating a deep similarity metric to prevent ID switches, handle occlusions, and propose new detections, achieving competitive results in real-time benchmarks.
Tracking by detection is a common approach to solving the Multiple Object Tracking problem. In this paper we show how learning a deep similarity metric can improve three key aspects of pedestrian tracking on a multiple object tracking benchmark. We train a convolutional neural network to learn an embedding function in a Siamese configuration on a large person re-identification dataset. The offline-trained embedding network is integrated in to the tracking formulation to improve performance while retaining real-time performance. The proposed tracker stores appearance metrics while detections are strong, using this appearance information to: prevent ID switches, associate tracklets through occlusion, and propose new detections where detector confidence is low. This method achieves competitive results in evaluation, especially among online, real-time approaches. We present an ablative study showing the impact of each of the three uses of our deep appearance metric.