Josh McDermott's lecture from MIT's 9.35 Perception course examines mid-level vision, the stage of processing that organizes raw visual signals into coherent surfaces, objects, and depth before higher cognitive interpretation takes over. McDermott builds on earlier lectures covering low-level vision, walking through how the brain groups edges and regions, infers boundaries between objects, and extracts cues like occlusion and shading to reconstruct a three-dimensional scene from a two-dimensional retinal image. The lecture situates mid-level vision within the broader hierarchical model of visual processing, showing how outputs from earlier stages feed into grouping and segmentation mechanisms. Drawn from MIT OpenCourseWare's Spring 2024 offering, the talk is aimed at students already familiar with basic sensory neuroscience and vision science terminology, and it uses diagrams and example images to illustrate how perceptual organization works before the visual system arrives at recognizable objects.