Alassane Watt

Visual Object Search by Learning Spatial Context

We present a visual navigation approach that uses context information to navigate an agent to find and reach a target object. To learn context from the objects present in the scene, we transform visual information into an intermediate representation called context grid which essentially represents h