Placing point-feature labels on interactive maps must balance geometric validity, display yield, local placement utility, and stability under camera motion. Accessibility and multilingual requirements alter label dimensions, yet many algorithms reduce evaluation to simple overlap counts. We introduce LABELSENSE-Pilot, a reproducible prototype that generates eight compass-direction candidates per feature, scores them with a multilayer perceptron over graph-context summaries, adds a previous-placement bonus, and selects a final layout via mixed-integer optimization. Every produced layout is verified for viewport containment, per-feature uniqueness, and pairwise clearance. Experiments use 2,500 airport coordinates and names across 155 countries, with country-grouped train/test splits and generated stressors for density, camera, text suffix, preference, and enlarged‑font scenarios. Across five random seeds, LABELSENSE-Pilot displayed 85.62% of labels, incurred 2.09% flicker, and had zero collisions. Compared to a handcrafted‑utility integer program, it sacrificed 1.43 percentage points of display while reducing flicker by 12.04 points. Enlarged‑box‑aware layouts yielded zero proxy violations, whereas standard geometry evaluated at 1.5× violated 52.57% of selected placements. These results establish an auditable engineering trade‑off rather than a direct improvement in human accessibility, multilingual usability, or preference. Official recent baselines and participant evidence remain required before submission.
Review