Auxiliary tasks to improve trip hazard affordance detection on construction sites

Trip hazards are a danger present on all construction sites and failure to identify them can lead to serious injury. The major challenge of detecting these trips is the large number of objects capable of becoming hazards. This challenge is exacerbated because these hazards are not entirely defined by their object type, for example tools on the ground are trippable but not when hung on a wall. We propose to train a CNN performing pixel-wise trip detection with three auxiliary tasks to help the CNN better infer scene geometric properties of trip hazards. Of the three approaches investigated pixel-wise ground plane estimation, pixel depth estimation and pixel height above ground plane estimation, the first approach allowed the trip detector to achieve a 11.1% increase in Trip IOU over earlier work. These new approaches make it plausible to deploy a robotic platform to perform trip hazard detection, and so potentially reduce the number of injuries on construction sites.