I.emotional_response_initial = frustration.
I.emotional_response_secondary = need to rethink everything.
I.project_signvision.fps_issue = blurry and disappointing output.
I.project_signvision.architecture_changes = multiple revisions.
I.use_ai_for_diagnosis = true.
I.prefer_compare_ai_responses = true.
I.core_principle = understand the system, not just generate code.
These reflections capture my mindset while building SignVisionn: frustration at early FPS problems, a habit of rethinking the whole approach, and a commitment to using AI as a diagnostic partner while always probing the why behind any solution.
I built SignVisionn, my browser‑based ASL alphabet recognition system, using a dual‑pipeline vision engine with MediaPipe, EfficientNetB0 exported to ONNX, and runtime optimizations for real‑time inference.
I use a dual‑pipeline vision engine using MediaPipe.
I apply One Euro filtering to smooth prediction outputs.
I added device profiling to prevent memory crashes on lower‑end devices.
I achieve 47 FPS on an 8th‑gen i7 using 3 Web Workers.