Context-Interactive CNN for Person Re-Identification

Abstract

Despite growing progresses in recent years, crossscenario person re-identification remains challenging, mainly due to the pedestrians commonly surrounded by highly-complex environment contexts. In reality, the human perception mechanism could adaptively find proper contextualized spatial-temporal clues towards pedestrian recognition. However, conventional methods fall short in adaptively leveraging the long-term spatialtemporal information due to ever-increasing computational cost. Moreover, CNN-based deep learning methods are hard to conduct optimization due to the non-differentiable property of the builtin context search operation. To ameliorate, this paper proposes a novel Context-Interactive CNN (CI-CNN) to dynamically find both spatial and temporal contexts by embedding multi-task Reinforcement Learning (MTRL). The CI-CNN streamlines the multi-task reinforcement learning by using an actor-critic agent to capture the temporal-spatial context simultaneously, which comprises a context-policy network and a context-critic network. The former network learns policies to determine the optimal spatial context region and temporal sequence range. Based on the inferred temporal-spatial cues, the latter one focuses on the identification task and provides feedback for the policy network. Thus, CI-CNN can simultaneously zoom in/out the perception field in spatial and temporal domain for the context interaction with the environment.

Existing System

Coherence Constrained Graph LSTM (CCGLSTM)

Proposed System

We propose to solve the cross-scenario person Re-ID problem by integrating novel actor-critic agent in reinforcement learning embedded CNN, which can dynamically find the contexts and telescoping perception field, and thus gives rise to intrinsic information for person identification. We reveal the relations between the context and pedestrian for cross-scenario Re-ID, and formulate the interaction between optimal person regions and their spatial-temporal context, which gives rise to simple yet effective learning policies for obtaining optimal context clues. We design a novel deep reinforcement learning based multi-task framework to learn the context-agent by collaborating with the context policy network and context critic network.

Architecture

BLOCK DIAGRAM