Searching a Lightweight Network Architecture for Thermal Infrared Pedestrian Tracking

TL;DR

An early attempt to search an optimal network architecture for TIR-PT automatically, employing single-bottom and dual-bottom cells as basic search units and incorporating eight operation candidates within the search space, resulting in a high-performance network architecture that is both parameter- and computation-efficient.

Abstract

Manually-designed network architectures for thermal infrared pedestrian tracking (TIR-PT) require substantial effort from human experts. AlexNet and ResNet are widely used as backbone networks in TIR-PT applications. However, these architectures were originally designed for image classification and object detection tasks, which are less complex than the challenges presented by TIR-PT. This paper makes an early attempt to search an optimal network architecture for TIR-PT automatically, employing single-bottom and dual-bottom cells as basic search units and incorporating eight operation candidates within the search space. To expedite the search process, a random channel selection strategy is employed prior to assessing operation candidates. Classification, batch hard triplet, and center loss are jointly used to retrain the searched architecture. The outcome is a high-performance network architecture that is both parameter- and computation-efficient. Extensive experiments proved the effectiveness of the automated method.

Figures (15)

• Figure 1: Overview of the proposed TIR-PT method.
• Figure 2: A single-bottom cell contains six nodes. Node S is the input feature map. Nodes 0, 1, 2, 3 are intermediate feature maps. The last output node is depth-wise concatenation of four intermediate nodes. Each blue edge denotes eight operation candidates.
• Figure 3: A dual-bottom cell contains seven nodes. Nodes S0 and S1 are feature maps outputted from the two previous layers.
• Figure 4: Architecture parameters $\alpha$ for a dual-bottom cell. Rows and columns represent 14 edges and 8 operation candidates, respectively.
• Figure 5: The overview of network architecture sample with only one input (a.k.a., single-bottom).