RTMPose-s LiteRT fp16 (fully-GPU, Pixel 8a corr 0.999, 4ms)

README.md

@@ -0,0 +1,56 @@
+---
+license: apache-2.0
+library_name: LiteRT
+pipeline_tag: keypoint-detection
+tags: [litert, tflite, on-device, android, gpu, pose-estimation, keypoint-detection, rtmpose, mmpose]
+base_model: open-mmlab/mmpose
+---
+
+# RTMPose-s — LiteRT (on-device real-time 2D human pose, fully-GPU)
+
+[RTMPose](https://github.com/open-mmlab/mmpose/tree/main/projects/rtmpose) (mmpose, CSPNeXt backbone +
+RTMCC/SimCC head) top-down 2D human pose, converted to **LiteRT** and running **fully on the `CompiledModel`
+GPU** (ML Drift) on Android. Estimates 17 COCO keypoints for a single centered person — the SOTA real-time
+pose model, device-verified end-to-end.
+
+![RTMPose-s — input | skeleton (on-device LiteRT GPU)](samples/sample.png)
+
+## On-device (Pixel 8a, Tensor G3 — verified)
+
+| | |
+|---|---|
+| nodes on GPU | **256 / 256** LITERT_CL (full residency) |
+| inference | **~4 ms** (256×192) |
+| size | 11.1 MB (fp16) |
+| accuracy | device-vs-PyTorch SimCC corr **0.999**, keypoints within **0.3 px** (max 1 px) |
+
+```
+image[1,3,256,192] (ImageNet 0-255 norm) →[GPU: CSPNeXt + RTMCC]→ simcc_x[1,17,384], simcc_y[1,17,512]
+```
+
+The SimCC head emits two 1D distributions per keypoint; argmax over the bins (÷ split=2) gives the pixel x/y.
+
+## How it converts (litert-torch) — two numerically-exact re-authorings
+
+Both are **on-device-only** Mali issues: they pass the desktop op-check and report full LITERT_CL residency,
+yet the device output was wrong until fixed (*residency ≠ correctness*):
+
+1. **`ScaleNorm` (RMS norm) fp16 overflow → all-zero head.** The RTMCC `ScaleNorm` input reaches ≈ |274|, so
+   its channel `Σ x²` ≈ 3.6M **overflows fp16 (max 65504)** on the Mali delegate (which reduces in fp16 even
+   for an fp32 graph) → `norm = ∞` → `x/∞ = 0` → the whole head collapses to zero. Fix: scale `x` down by
+   S=64 **before** squaring, then rescale (math-identical) — a SafeRMSNorm.
+2. **GAU attention `act@act` BMM → broadcast-reduce.** The Gated Attention Unit's `q@kᵀ` and `kernel@v` are
+   activation×activation batch-matmuls that the Mali delegate mis-computes; at K=17 tokens the exact
+   replacement is `(q[:,:,None,:]·k[:,None,:,:]).sum(-1)`.
+
+Result: banned ops NONE, all tensors ≤4D, tflite-vs-torch corr **1.0**, device-vs-torch corr **0.999**.
+
+## Preprocessing
+
+Center-crop to 3:4, resize to 192×256, ImageNet 0-255 normalize (mean [123.675, 116.28, 103.53], std
+[58.395, 57.12, 57.375]), NCHW planar. Top-down — expects one roughly-centered person.
+
+## License
+
+[Apache-2.0](https://github.com/open-mmlab/mmpose/blob/main/LICENSE). Upstream:
+[open-mmlab/mmpose](https://github.com/open-mmlab/mmpose) RTMPose-s.

rtmpose_s_fp16.tflite

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89f2d5c921bd824a39b37fc718550909cbb60d90e1cba48aa88b3ef71fab631d
+size 11139472