ncnn 0.1.0

ncnn (Vulkan) inference bindings for Dart/Flutter — a thin C shim over ncnn::Net exposed via dart:ffi. Runs image inference on CPU or any Vulkan-capable GPU (NVIDIA / AMD / Intel / Apple via MoltenVK [...]

0.1.0

Initial release.

• ncnn (Vulkan) inference bindings for Dart/Flutter via a thin C shim and dart:ffi — CPU + any Vulkan GPU (NVIDIA / AMD / Intel / Apple via MoltenVK / Android), automatic CPU fallback.
• Five-platform FFI plugin (Android / iOS / Linux / macOS / Windows); official ncnn prebuilts downloaded at build time.
• NcnnNet: shape-driven multi-output inference with capacity retry, configurable preprocessing (mean/norm), pixel format and input blob (ultralytics YOLO conventions as defaults).
• hn_extract_f32: raw float tensor input for non-image models.
• NcnnInferenceEngine: backend selection, best-device heuristic, GPU→CPU fallback, injectable backends for testing.
• Windows ncnn_helper child process: full GPU acceleration outside the Flutter engine process (ncnn's Vulkan init crashes inside it — dump-verified; see README).
• YOLO post-processing utilities: classify (argmax/topK/softmax), detect (grid decode + per-class NMS).