Upload 6 files

README.md

@@ -109,6 +109,64 @@ with torch.no_grad():
 | 4 | 96.83% | 0.9600 |
 | **5** | **97.35%** | **0.9666** |
 
+## Files
+
+| File | Size | Description |
+|------|:----:|-------------|
+| `pytorch_model.bin` | 121 MB | PyTorch weights (FP32) |
+| `model.onnx` | 164 MB | ONNX model for mobile deployment |
+| `config.json` | 1.2 KB | Model configuration |
+| `model.py` | 6.9 KB | Model architecture code |
+| `convert_coreml_macos.py` | 2.2 KB | CoreML conversion script (macOS) |
+
+## Platform-specific Usage
+
+### PyTorch (Server/Desktop)
+
+```python
+import torch
+from model import DriverBehaviorModel
+
+model = DriverBehaviorModel(num_classes=5, pretrained=False)
+checkpoint = torch.load("pytorch_model.bin", map_location="cpu")
+model.load_state_dict(checkpoint["model"])
+model.eval()
+```
+
+### iOS (CoreML)
+
+1. Copy `model.onnx` to macOS
+2. Run conversion script:
+```bash
+python convert_coreml_macos.py
+```
+3. Add generated `DriverBehavior.mlpackage` to Xcode project
+
+### Android (ONNX Runtime)
+
+```kotlin
+// build.gradle
+implementation 'com.microsoft.onnxruntime:onnxruntime-android:1.16.0'
+
+// Kotlin
+val session = OrtEnvironment.getEnvironment()
+    .createSession(assetManager.open("model.onnx").readBytes())
+
+val output = session.run(mapOf("video_input" to inputTensor))
+```
+
+## Preprocessing (All Platforms)
+
+```
+Input Shape: [1, 3, 30, 224, 224]  (batch, channels, frames, height, width)
+Channel Order: RGB
+Normalization: (pixel / 255.0 - mean) / std
+  - mean = [0.485, 0.456, 0.406]
+  - std = [0.229, 0.224, 0.225]
+Resize: 224x224 (BILINEAR)
+Frames: 30 frames uniformly sampled
+```
+
 ## License
 
 This model is for research purposes only.

convert_coreml_macos.py

@@ -0,0 +1,84 @@
+#!/usr/bin/env python3
+"""
+macOS에서 실행: ONNX → CoreML FP16 변환
+
+사용법:
+    python convert_coreml_macos.py
+
+필요 패키지:
+    pip install coremltools onnx
+"""
+
+import coremltools as ct
+import numpy as np
+
+# 입력 파일
+ONNX_PATH = "model.onnx"  # 같은 폴더에 있어야 함
+OUTPUT_PATH = "DriverBehavior.mlpackage"
+
+print("=" * 60)
+print("ONNX → CoreML FP16 변환 (macOS)")
+print("=" * 60)
+
+# 변환
+print("\n[1] CoreML 변환 중...")
+mlmodel = ct.convert(
+    ONNX_PATH,
+    minimum_deployment_target=ct.target.iOS15,
+    convert_to="mlprogram",
+    compute_precision=ct.precision.FLOAT16,  # FP16 (Metal 최적화)
+    compute_units=ct.ComputeUnit.ALL,  # CPU + GPU + Neural Engine
+)
+
+# 메타데이터
+mlmodel.author = "C-Team"
+mlmodel.short_description = "Driver Behavior Detection - Video Swin Transformer"
+mlmodel.version = "1.0"
+
+# 저장
+mlmodel.save(OUTPUT_PATH)
+print(f"✓ 저장 완료: {OUTPUT_PATH}")
+
+# 검증
+print("\n[2] 모델 검증 중...")
+import onnxruntime as ort
+
+# 더미 입력
+dummy_input = np.random.randn(1, 3, 30, 224, 224).astype(np.float32)
+
+# ONNX 출력
+sess = ort.InferenceSession(ONNX_PATH)
+onnx_out = sess.run(None, {'video_input': dummy_input})[0]
+
+# CoreML 출력
+coreml_out = mlmodel.predict({'video_input': dummy_input})['class_logits']
+
+# 비교
+max_diff = np.abs(onnx_out - coreml_out).max()
+print(f"  ONNX vs CoreML 최대 차이: {max_diff:.6f}")
+
+onnx_pred = onnx_out.argmax()
+coreml_pred = coreml_out.argmax()
+
+if onnx_pred == coreml_pred:
+    print("  ✓ 예측 클래스 일치!")
+else:
+    print(f"  ⚠ 예측 다름: ONNX={onnx_pred}, CoreML={coreml_pred}")
+
+print("\n" + "=" * 60)
+print("변환 완료!")
+print("=" * 60)
+print(f"""
+iOS 사용법:
+1. {OUTPUT_PATH}를 Xcode 프로젝트에 드래그앤드롭
+2. 자동으로 Metal FP16 가속 적용됨
+
+입력 형식:
+- Shape: [1, 3, 30, 224, 224]
+- 정규화: (pixel/255 - [0.485,0.456,0.406]) / [0.229,0.224,0.225]
+- 채널: RGB
+
+출력:
+- [1, 5] 로짓 → argmax로 클래스 예측
+- 0: 정상, 1: 졸음운전, 2: 물건찾기, 3: 휴대폰사용, 4: 운전자폭행
+""")

model.onnx

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:036ac05c4b782d2ea1e24eb61366f197e0692ba24abaddca6798d56c1a337cec
+size 171169182