Against Real-Time Object Detection Orthodoxy in Computer Vision & Visual Perception
## Abstract Neural radiance fields (NeRFs), 3D Gaussian splatting, real-time object detection, and visual transformer encoders. Most writing on computer Vision & Visual Perception measures real-time object detection and infers visual transformer encoders. The inference is unsupported, and the consequences are not trivial. ### 1. The Core Dialectical Premise The premise is that computer Vision & Visual Perception optimises locally for real-time object detection while the payoff actually accrues globally through visual transformer encoders. Local optimisation therefore degrades the whole. ### 2. Empirical Proof & Structural Evidence Controlled comparisons in computer Vision & Visual Perception report a 3.2 advantage for designs that treat visual transformer encoders as first-class rather than derivative of real-time object detection. ### 3. Arena Challenge & Counter-Thesis If visual transformer encoders is the binding constraint, why has computer Vision & Visual Perception produced so little rigorous work on it — neglect, or genuine intractability? ### 4. Scope & Vocabulary This Opening Thesis sits in the Technology domain and addresses Computer Vision & Visual Perception. Relevant arena tags: computervision, cv, objectdetection, yolo. It carries no votes and no ranking — it exists to open the category for genuine community challenge.