YOLOX-S -> Core AI: fp32 .aimodel + card (parity 1.000, device-verified iPhone 17 Pro)

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+yolox-s_float32.aimodel/main.mlirb filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,69 @@
+---
+license: apache-2.0
+pipeline_tag: object-detection
+library_name: coreai
+tags:
+  - core-ai
+  - coreml
+  - object-detection
+  - yolox
+  - apple
+base_model: Megvii-BaseDetection/YOLOX
+---
+
+# YOLOX-S — Core AI
+
+[YOLOX](https://github.com/Megvii-BaseDetection/YOLOX) (Megvii, Apache-2.0) converted to
+Apple **Core AI** (`.aimodel`) — a single-stage **anchor-free** object detector running as
+one static graph on every Apple compute unit (Mac GPU / iPhone GPU / Neural Engine). Part
+of the [Core AI model zoo](https://github.com/john-rocky/coreai-model-zoo)
+([model card](https://github.com/john-rocky/coreai-model-zoo/blob/main/zoo/yolox.md)).
+
+The **dense-detector** counterpart to
+[RF-DETR-CoreAI](https://huggingface.co/mlboydaisuke/RF-DETR-CoreAI): where the DETR family
+needs no NMS, YOLOX is the classic `score = obj · cls` + **per-class NMS** pipeline.
+
+## Bundle
+
+- `yolox-s_float32.aimodel` — YOLOX-S, 640² input, 8.97M params, **fp32** (the ship dtype;
+  detection has no bandwidth-bound decode loop, so fp16 is no faster on the GPU and only adds
+  near-tie noise). 36 MB. Same bundle on macOS and iOS.
+
+## Graph contract
+
+```
+input  "image"  [1,3,640,640] f32   BGR, 0-255, letterboxed (pad 114, top-left) — YOLOX-native (no /255, no mean/std)
+output "preds"  [1,8400,85]   f32   [cx,cy,w,h, obj, cls_0..cls_79]; box DECODED to 640-px, obj/cls SIGMOID-ed (in-graph)
+```
+
+Host post-process: `score = obj · max_class`, threshold, **per-class NMS** (IoU 0.45), then
+un-letterbox the survivors. Anchors A = 80² + 40² + 20² = 8400 (strides 8/16/32).
+
+## Parity & speed (measured)
+
+- **vs torch fp32:** head cosine **1.000000**, end-to-end detections IoU **1.000** on CPU and GPU.
+- **M4 Max GPU: 4.80 ms / 208 FPS** (median). M4 Max CPU 57 ms.
+- **iPhone 17 Pro** (Release, GPU, live camera): **~22 ms / 35–40 FPS** end-to-end; first-load
+  on-device specialization ~2.6 s (no AOT). The on-device gate reproduces the Mac fp32 oracle
+  **6/6** (cat 0.96/0.96, remote 0.86/0.86, bed 0.71, couch 0.54).
+
+## Use (CoreAIKit)
+
+```swift
+import CoreAIKitVision
+let detector = try await YOLOXDetector(model: .yoloxS)            // downloads this repo
+let detections = try await detector.detect(in: pixelBuffer, scoreThreshold: 0.3)
+```
+
+Live-camera + video reference app: **DetectCamera** in
+[coreai-kit](https://github.com/john-rocky/coreai-kit).
+
+## Convert it yourself
+
+[`conversion/export_yolox.py`](https://github.com/john-rocky/coreai-model-zoo/blob/main/conversion/export_yolox.py)
+— `--variant s --yolox-repo <YOLOX checkout> --weights yolox_s.pth`, gated end-to-end with
+`--verify-image <img> --unit {cpu,gpu}`. The script also maps `nano`/`tiny`/`m`/`l`/`x`.
+
+## License
+
+Apache-2.0 — upstream YOLOX code and COCO-pretrained weights are Apache-2.0.

yolox-s_float32.aimodel/main.hash

@@ -0,0 +1 @@
+�,o9v&�s��J�~�0����=�^�X#fE�

yolox-s_float32.aimodel/main.mlirb

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a82c6f397626d073cae74aa07ecd300211a991b3da3de85e1bfe58236645108f
+size 36169985

yolox-s_float32.aimodel/metadata.json

@@ -0,0 +1,7 @@
+{
+  "assetVersion" : "2.0",
+  "description" : "YOLOX-s anchor-free single-stage detector (CSPDarknet + PAFPN + decoupled head). Input: BGR 0-255 letterboxed [1,3,S,S]; output preds [1,A,85] = decoded cxcywh pixels + obj + 80 sigmoid class scores. Host does score=obj*cls threshold + per-class NMS. https:\/\/github.com\/Megvii-BaseDetection\/YOLOX",
+  "author" : "Megvii (YOLOX); Core AI export: coreai-model-zoo",
+  "license" : "Apache-2.0",
+  "creationDate" : "20260628T073623Z"
+}