The Neurorobotics Platform of the Human Brain Project hosts many different large-scale models that can easily be connected with each other. Here, we linked a deep neural network for saliency computation to a spiking cortical model for visual segmentation (the Laminart model) to investigate global effects in visual crowding. Global effects are observed, in a Vernier discrimination task, where the target is flanked by one to seven squares spanning up to seventeen degrees of the peripheral visual field. Contrary to predictions of most models, crowding is lower as more squares flank the target. In the simulation, the saliency model uses fast bottom-up information to bias the segmentation processes of the Laminart model towards specific parts of the visual stimulus. Simulations of the model with the same stimuli as in empirical studies produce essentially the same behavior as human observers.