Upload 6 files

LICENSE

@@ -0,0 +1,190 @@
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to the Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   Copyright 2025 C-Team
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

README.md

@@ -1,3 +1,303 @@
+# BaramNuri (바람누리) - Driver Behavior Detection Model
+
+<div align="center">
+
+**바람누리** | *Wind that watches over the world*
+
+경량화된 운전자 이상행동 탐지 AI 모델
+
+[![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE)
+[![Python](https://img.shields.io/badge/Python-3.8+-green.svg)](https://python.org)
+[![PyTorch](https://img.shields.io/badge/PyTorch-2.0+-red.svg)](https://pytorch.org)
+
+</div>
+
 ---
-license: apache-2.0
+
+## Model Description
+
+**바람누리(BaramNuri)**는 차량 내 카메라 영상에서 운전자의 이상행동을 실시간으로 탐지하는 경량화 딥러닝 모델입니다.
+
+### Key Features
+
+- **경량화**: Teacher 모델(27.86M) 대비 **49% 파라미터 감소** (14.20M)
+- **고성능**: Knowledge Distillation으로 **98% 성능 유지**
+- **실시간**: 엣지 디바이스 배포 가능 (INT8: ~13MB)
+- **5종 분류**: 정상, 졸음운전, 물건찾기, 휴대폰 사용, 운전자 폭행
+
 ---
+
+## Architecture
+
+```
+┌─────────────────────────────────────────────────────────────────┐
+│                     BaramNuri Architecture                       │
+├─────────────────────────────────────────────────────────────────┤
+│                                                                  │
+│   Input: [B, 3, 30, 224, 224] (1초 영상, 30fps)                 │
+│                         │                                        │
+│                         ▼                                        │
+│   ┌─────────────────────────────────────┐                       │
+│   │     Video Swin-T (Stage 1-3)        │  ← Kinetics-400       │
+│   │     Shifted Window Attention         │     Pretrained        │
+│   │     Output: 384 dim features        │                       │
+│   └─────────────────────────────────────┘                       │
+│                         │                                        │
+│                         ▼                                        │
+│   ┌─────────────────────────────────────┐                       │
+│   │     Selective SSM Block (x2)        │  ← Mamba-style        │
+│   │     - 1D Conv for local context     │     Temporal          │
+│   │     - Selective state space         │     Modeling          │
+│   │     - Input-dependent B, C, delta   │                       │
+│   └─────────────────────────────────────┘                       │
+│                         │                                        │
+│                         ▼                                        │
+│   ┌─────────────────────────────────────┐                       │
+│   │     Classification Head             │                       │
+│   │     LayerNorm → Dropout → Linear    │                       │
+│   └─────────────────────────────────────┘                       │
+│                         │                                        │
+│                         ▼                                        │
+│   Output: [B, 5] (5-class logits)                               │
+│                                                                  │
+└─────────────────────────────────────────────────────────────────┘
+```
+
+### Why This Architecture?
+
+| Component | Purpose | Benefit |
+|-----------|---------|---------|
+| **Video Swin (Stage 1-3)** | Spatial feature extraction | Proven performance on video |
+| **Stage 4 Removal** | 55% parameter reduction | Lightweight without quality loss |
+| **Selective SSM** | Temporal modeling | O(n) complexity vs O(n²) attention |
+| **Knowledge Distillation** | Performance retention | Learn from larger teacher model |
+
+---
+
+## Performance
+
+### Classification Metrics
+
+| Metric | Score |
+|--------|-------|
+| **Accuracy** | 96.17% |
+| **Macro F1** | 0.9504 |
+| **Precision** | 0.95 |
+| **Recall** | 0.95 |
+
+### Per-Class Performance
+
+| Class | Precision | Recall | F1-Score |
+|-------|:---------:|:------:|:--------:|
+| 정상 (Normal) | 0.93 | 0.93 | 0.93 |
+| 졸음운전 (Drowsy) | 0.98 | 0.97 | 0.97 |
+| 물건찾기 (Searching) | 0.93 | 0.95 | 0.94 |
+| 휴대폰 사용 (Phone) | 0.94 | 0.93 | 0.94 |
+| 운전자 폭행 (Assault) | 0.99 | 0.99 | 0.99 |
+
+### Comparison with Teacher
+
+| Metric | Teacher | BaramNuri | Comparison |
+|--------|---------|-----------|------------|
+| **Parameters** | 27.86M | 14.20M | **-49%** |
+| **Model Size (FP32)** | ~106 MB | ~54 MB | **-49%** |
+| **Model Size (INT8)** | ~26 MB | ~13 MB | **-50%** |
+| **Accuracy** | 98.05% | 96.17% | 98.1% retained |
+| **Macro F1** | 0.9757 | 0.9504 | 97.4% retained |
+
+---
+
+## Quick Start
+
+### Installation
+
+```bash
+pip install torch torchvision
+```
+
+### Inference
+
+```python
+import torch
+from model import BaramNuri
+
+# Load model
+model = BaramNuri(num_classes=5, pretrained=False)
+checkpoint = torch.load('baramnuri_beta.pth', map_location='cpu')
+model.load_state_dict(checkpoint['model_state_dict'])
+model.eval()
+
+# Prepare input (1 second video, 30fps, 224x224)
+# Shape: [batch, channels, frames, height, width]
+video = torch.randn(1, 3, 30, 224, 224)
+
+# Inference
+with torch.no_grad():
+    logits = model(video)
+    probs = torch.softmax(logits, dim=-1)
+    pred_class = probs.argmax(dim=-1).item()
+
+# Class names
+class_names = ["정상", "졸음운전", "물건찾기", "휴대폰 사용", "운전자 폭행"]
+print(f"Predicted: {class_names[pred_class]} ({probs[0, pred_class]:.2%})")
+```
+
+### With Prediction Helper
+
+```python
+# Single prediction with confidence
+result = model.predict(video)
+print(f"Class: {result['class_name']}")
+print(f"Confidence: {result['confidence']:.2%}")
+```
+
+---
+
+## Input Specification
+
+| Parameter | Value |
+|-----------|-------|
+| **Format** | `[B, C, T, H, W]` (BCTHW) |
+| **Channels** | 3 (RGB) |
+| **Frames** | 30 (1 second at 30fps) |
+| **Resolution** | 224 x 224 |
+| **Normalization** | ImageNet mean/std |
+
+### Preprocessing
+
+```python
+from torchvision import transforms
+
+transform = transforms.Compose([
+    transforms.Resize((224, 224)),
+    transforms.ToTensor(),
+    transforms.Normalize(
+        mean=[0.485, 0.456, 0.406],
+        std=[0.229, 0.224, 0.225]
+    ),
+])
+```
+
+---
+
+## Training Details
+
+### Knowledge Distillation
+
+```
+Teacher: Video Swin-T (27.86M, 98.05% acc)
+    │
+    │  Soft Labels (Temperature=4.0)
+    ▼
+Student: BaramNuri (14.20M)
+    │
+    │  L = 0.5 * L_hard + 0.5 * L_soft
+    ▼
+Result: 96.17% acc (98% of teacher performance)
+```
+
+### Training Configuration
+
+| Parameter | Value |
+|-----------|-------|
+| Optimizer | AdamW |
+| Learning Rate | 1e-4 |
+| Weight Decay | 0.05 |
+| Batch Size | 96 (effective) |
+| Epochs | 6 |
+| Loss | CE + KL Divergence |
+| Temperature | 4.0 |
+| Alpha (hard/soft) | 0.5 |
+
+---
+
+## Deployment
+
+### Server Deployment (GPU)
+
+```python
+model = BaramNuri(num_classes=5)
+model.load_state_dict(torch.load('baramnuri_beta.pth')['model_state_dict'])
+model = model.cuda().eval()
+
+# FP16 for faster inference
+model = model.half()
+```
+
+### Edge Deployment (INT8 Quantization)
+
+```python
+import torch.quantization as quant
+
+model_int8 = quant.quantize_dynamic(
+    model, {torch.nn.Linear}, dtype=torch.qint8
+)
+# Model size: ~13MB
+```
+
+### ONNX Export
+
+```python
+dummy_input = torch.randn(1, 3, 30, 224, 224)
+torch.onnx.export(
+    model, dummy_input, "baramnuri.onnx",
+    input_names=['video'],
+    output_names=['logits'],
+    dynamic_axes={'video': {0: 'batch'}}
+)
+```
+
+---
+
+## Use Cases
+
+1. **Fleet Management**: Monitor driver behavior in commercial vehicles
+2. **Insurance Telematics**: Risk assessment based on driving behavior
+3. **ADAS Integration**: Advanced driver assistance systems
+4. **Safety Research**: Analyze driving patterns and fatigue
+
+---
+
+## Limitations
+
+- Trained on Korean driving environment data
+- Requires frontal camera facing the driver
+- Optimal performance at 30fps input
+- May require fine-tuning for different camera angles
+
+---
+
+## Citation
+
+```bibtex
+@misc{baramnuri2025,
+  title={BaramNuri: Lightweight Driver Behavior Detection with Knowledge Distillation},
+  author={C-Team},
+  year={2025},
+  howpublished={\url{https://huggingface.co/c-team/baramnuri-beta}}
+}
+```
+
+---
+
+## License
+
+This model is released under the [Apache 2.0 License](LICENSE).
+
+---
+
+## Acknowledgments
+
+- Video Swin Transformer: Liu et al. (CVPR 2022)
+- Knowledge Distillation: Hinton et al. (2015)
+- Mamba/S4: Gu & Dao (2023)
+
+---
+
+<div align="center">
+
+**바람누리** - 안전한 운전을 위한 AI
+
+Made with care by C-Team
+
+</div>

baramnuri_beta.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb367c1643c1d53b08f703224f6174fb336ece611a0c4ca295e41befa9aca760
+size 182925595

config.json

@@ -0,0 +1,92 @@
+{
+  "model_type": "baramnuri",
+  "architectures": ["BaramNuri"],
+  "model_name": "baramnuri-beta",
+  "version": "0.1.0-beta",
+
+  "num_classes": 5,
+  "class_names": ["정상", "졸음운전", "물건찾기", "휴대폰 사용", "운전자 폭행"],
+  "class_names_en": ["normal", "drowsy_driving", "searching_object", "phone_usage", "driver_assault"],
+
+  "backbone": {
+    "type": "video_swin_t",
+    "pretrained_on": "kinetics-400",
+    "stages_used": [1, 2, 3],
+    "feature_dim": 384
+  },
+
+  "ssm_block": {
+    "type": "selective_ssm",
+    "d_state": 16,
+    "d_conv": 4,
+    "expand": 2,
+    "n_layers": 2,
+    "dropout": 0.2
+  },
+
+  "input_spec": {
+    "channels": 3,
+    "num_frames": 30,
+    "height": 224,
+    "width": 224,
+    "fps": 30,
+    "format": "BCTHW"
+  },
+
+  "model_stats": {
+    "total_parameters": 14203205,
+    "total_parameters_readable": "14.20M",
+    "model_size_fp32_mb": 54,
+    "model_size_fp16_mb": 27,
+    "model_size_int8_mb": 13
+  },
+
+  "training": {
+    "method": "knowledge_distillation",
+    "teacher_model": "Video Swin-T (27.86M)",
+    "teacher_accuracy": 0.9805,
+    "teacher_f1": 0.9757,
+    "epochs_trained": 6,
+    "best_accuracy": 0.9617,
+    "best_macro_f1": 0.9504,
+    "optimizer": "AdamW",
+    "learning_rate": 1e-4,
+    "weight_decay": 0.05,
+    "batch_size": 96,
+    "data_augmentation": ["resize", "normalize"]
+  },
+
+  "performance": {
+    "accuracy": 0.9617,
+    "macro_f1": 0.9504,
+    "per_class_f1": {
+      "정상": 0.93,
+      "졸음운전": 0.97,
+      "물건찾기": 0.94,
+      "휴대폰 사용": 0.94,
+      "운전자 폭행": 0.99
+    }
+  },
+
+  "comparison_with_teacher": {
+    "parameter_reduction": "49%",
+    "size_reduction": "49%",
+    "accuracy_retention": "98.1%",
+    "f1_retention": "97.4%",
+    "training_speed_improvement": "40%"
+  },
+
+  "license": "Apache-2.0",
+  "language": ["ko", "en"],
+  "tags": [
+    "video-classification",
+    "driver-behavior",
+    "knowledge-distillation",
+    "video-swin-transformer",
+    "state-space-model",
+    "ssm",
+    "mamba-style",
+    "lightweight",
+    "korean"
+  ]
+}

model.py

@@ -0,0 +1,365 @@
+"""
+BaramNuri (바람누리) - Lightweight Driver Behavior Detection Model
+
+A hybrid architecture combining:
+- Video Swin Transformer (Stage 1-3) for spatial features
+- Selective State Space Model (SSM) for temporal modeling
+
+Trained via Knowledge Distillation from Video Swin-T teacher.
+
+Author: C-Team
+License: Apache-2.0
+"""
+
+import torch
+import torch.nn as nn
+import torch.nn.functional as F
+from torchvision.models.video import swin3d_t, Swin3D_T_Weights
+from typing import Dict, Tuple
+
+
+class SelectiveSSM(nn.Module):
+    """
+    Selective State Space Model (Mamba-style)
+
+    Key: Dynamically generates B, C, delta based on input
+    - Important information is remembered
+    - Less important information is quickly forgotten
+    """
+
+    def __init__(self, d_model: int, d_state: int = 16, d_conv: int = 4, expand: int = 2, dropout: float = 0.1):
+        super().__init__()
+
+        self.d_model = d_model
+        self.d_state = d_state
+        self.d_conv = d_conv
+        self.expand = expand
+        self.d_inner = d_model * expand
+
+        # Input projection (expansion)
+        self.in_proj = nn.Linear(d_model, self.d_inner * 2, bias=False)
+
+        # 1D convolution (local context)
+        self.conv1d = nn.Conv1d(
+            self.d_inner, self.d_inner,
+            kernel_size=d_conv,
+            padding=d_conv - 1,
+            groups=self.d_inner
+        )
+
+        # SSM parameter generator (selective!)
+        self.x_proj = nn.Linear(self.d_inner, d_state * 2 + 1, bias=False)
+
+        # A parameter (learnable diagonal matrix)
+        self.A_log = nn.Parameter(torch.log(torch.arange(1, d_state + 1, dtype=torch.float32)))
+        self.D = nn.Parameter(torch.ones(self.d_inner))
+
+        # Output projection
+        self.out_proj = nn.Linear(self.d_inner, d_model, bias=False)
+
+        self.dropout = nn.Dropout(dropout)
+        self.layer_norm = nn.LayerNorm(d_model)
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        """
+        Args:
+            x: [B, T, D]
+        Returns:
+            y: [B, T, D]
+        """
+        residual = x
+        x = self.layer_norm(x)
+
+        B, T, D = x.shape
+
+        # Input projection -> [B, T, 2*d_inner]
+        xz = self.in_proj(x)
+        x, z = xz.chunk(2, dim=-1)
+
+        # 1D Conv (capture local context)
+        x = x.transpose(1, 2)
+        x = self.conv1d(x)[:, :, :T]
+        x = x.transpose(1, 2)
+
+        x = F.silu(x)
+
+        # Selective SSM parameter generation
+        x_ssm = self.x_proj(x)
+        B_t = x_ssm[:, :, :self.d_state]
+        C_t = x_ssm[:, :, self.d_state:self.d_state*2]
+        delta = F.softplus(x_ssm[:, :, -1:])
+
+        # A parameter (negative for stability)
+        A = -torch.exp(self.A_log)
+
+        # Discretization: A_bar = exp(delta * A)
+        A_bar = torch.exp(delta * A.view(1, 1, -1))
+
+        # SSM scan
+        h = torch.zeros(B, self.d_inner, self.d_state, device=x.device, dtype=x.dtype)
+        outputs = []
+
+        for t in range(T):
+            x_t = x[:, t, :]
+            B_t_t = B_t[:, t, :]
+            C_t_t = C_t[:, t, :]
+            A_bar_t = A_bar[:, t, :]
+
+            # h = A_bar * h + B_t * x
+            h = h * A_bar_t.unsqueeze(1) + B_t_t.unsqueeze(1) * x_t.unsqueeze(-1)
+
+            # y = C_t * h + D * x
+            y_t = (C_t_t.unsqueeze(1) * h).sum(dim=-1) + self.D * x_t
+            outputs.append(y_t)
+
+        y = torch.stack(outputs, dim=1)
+
+        # Gating
+        y = y * F.silu(z)
+
+        # Output projection
+        y = self.out_proj(y)
+        y = self.dropout(y)
+
+        return y + residual
+
+
+class TemporalSSMBlock(nn.Module):
+    """
+    Temporal SSM Block for video
+
+    Takes [B, T, C] sequence and applies SSM layers
+    """
+
+    def __init__(self, d_model: int, d_state: int = 16, n_layers: int = 2, dropout: float = 0.1):
+        super().__init__()
+
+        self.ssm_layers = nn.ModuleList([
+            SelectiveSSM(d_model, d_state=d_state, dropout=dropout)
+            for _ in range(n_layers)
+        ])
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        """
+        Args:
+            x: [B, T, D] sequence
+        Returns:
+            y: [B, D] final representation
+        """
+        for ssm in self.ssm_layers:
+            x = ssm(x)
+
+        return x.mean(dim=1)
+
+
+class BaramNuri(nn.Module):
+    """
+    BaramNuri (바람누리) - Lightweight Driver Behavior Detection Model
+
+    Architecture:
+    1. Video Swin-T Stages 1-3 (spatial features, 384 dim)
+    2. Selective SSM Block (temporal modeling)
+    3. Classification Head
+
+    Parameters: 14.20M (49% reduction from teacher's 27.86M)
+    Performance: 96.17% accuracy, 0.9504 Macro F1
+    """
+
+    CLASS_NAMES = ["정상", "졸음운전", "물건찾기", "휴대폰 사용", "운전자 폭행"]
+    CLASS_NAMES_EN = ["normal", "drowsy_driving", "searching_object", "phone_usage", "driver_assault"]
+
+    def __init__(
+        self,
+        num_classes: int = 5,
+        pretrained: bool = True,
+        d_state: int = 16,
+        ssm_layers: int = 2,
+        dropout: float = 0.2,
+    ):
+        super().__init__()
+
+        self.num_classes = num_classes
+
+        # Load Video Swin-T backbone (only Stage 1-3)
+        if pretrained:
+            print("Loading Swin backbone (Kinetics-400 pretrained)...")
+            full_swin = swin3d_t(weights=Swin3D_T_Weights.KINETICS400_V1)
+        else:
+            full_swin = swin3d_t(weights=None)
+
+        # Patch embedding
+        self.patch_embed = full_swin.patch_embed
+
+        # Use only Stage 1-3 (features[0:5]) for 384 dim output
+        self.features = nn.Sequential(*[full_swin.features[i] for i in range(5)])
+
+        # Stage 3 output: 384 dim
+        self.feature_dim = 384
+
+        # Global average pooling
+        self.avgpool = nn.AdaptiveAvgPool3d(output_size=1)
+
+        # SSM temporal modeling block
+        self.temporal_ssm = TemporalSSMBlock(
+            d_model=self.feature_dim,
+            d_state=d_state,
+            n_layers=ssm_layers,
+            dropout=dropout,
+        )
+
+        # Classification head
+        self.head = nn.Sequential(
+            nn.LayerNorm(self.feature_dim),
+            nn.Dropout(p=dropout),
+            nn.Linear(self.feature_dim, num_classes),
+        )
+
+        # Initialize head
+        self._init_head()
+
+        # Delete Stage 4 parameters (memory saving)
+        del full_swin
+
+    def _init_head(self):
+        """Initialize head weights"""
+        for m in self.head.modules():
+            if isinstance(m, nn.Linear):
+                nn.init.trunc_normal_(m.weight, std=0.02)
+                if m.bias is not None:
+                    nn.init.zeros_(m.bias)
+
+    def extract_features(self, x: torch.Tensor) -> torch.Tensor:
+        """
+        Extract features (for knowledge distillation)
+
+        Args:
+            x: [B, C, T, H, W]
+        Returns:
+            features: [B, feature_dim]
+        """
+        # Patch embedding
+        x = self.patch_embed(x)
+
+        # Swin Stages
+        x = self.features(x)
+
+        B, T, H, W, C = x.shape
+
+        # Spatial average -> [B, T, C] sequence
+        x = x.mean(dim=[2, 3])
+
+        # SSM temporal modeling
+        x = self.temporal_ssm(x)
+
+        return x
+
+    def forward(self, x: torch.Tensor) -> torch.Tensor:
+        """
+        Forward pass
+
+        Args:
+            x: [B, C, T, H, W] video tensor
+        Returns:
+            logits: [B, num_classes]
+        """
+        features = self.extract_features(x)
+        logits = self.head(features)
+        return logits
+
+    def forward_with_features(self, x: torch.Tensor) -> Tuple[torch.Tensor, torch.Tensor]:
+        """
+        Return both features and logits (for knowledge distillation)
+        """
+        features = self.extract_features(x)
+        logits = self.head(features)
+        return logits, features
+
+    def predict(self, x: torch.Tensor, return_english: bool = False) -> Dict:
+        """
+        Inference prediction
+
+        Args:
+            x: [1, C, T, H, W] single video
+            return_english: Return English class names
+        Returns:
+            dict with class, confidence, class_name
+        """
+        self.eval()
+        with torch.no_grad():
+            logits = self.forward(x)
+            probs = F.softmax(logits, dim=-1)[0]
+            class_idx = probs.argmax().item()
+
+            class_names = self.CLASS_NAMES_EN if return_english else self.CLASS_NAMES
+
+            return {
+                "class": class_idx,
+                "confidence": probs[class_idx].item(),
+                "class_name": class_names[class_idx],
+                "all_probs": {
+                    name: probs[i].item()
+                    for i, name in enumerate(class_names)
+                }
+            }
+
+    @classmethod
+    def from_pretrained(cls, checkpoint_path: str, device: str = 'cpu'):
+        """
+        Load pretrained model from checkpoint
+
+        Args:
+            checkpoint_path: Path to .pth file
+            device: 'cpu' or 'cuda'
+        Returns:
+            Loaded model in eval mode
+        """
+        model = cls(num_classes=5, pretrained=True)
+        checkpoint = torch.load(checkpoint_path, map_location=device)
+
+        if 'model_state_dict' in checkpoint:
+            model.load_state_dict(checkpoint['model_state_dict'])
+        else:
+            model.load_state_dict(checkpoint)
+
+        model = model.to(device)
+        model.eval()
+
+        return model
+
+
+def count_parameters(model: nn.Module) -> int:
+    """Count total model parameters"""
+    return sum(p.numel() for p in model.parameters())
+
+
+if __name__ == "__main__":
+    print("=" * 60)
+    print("BaramNuri Model Test")
+    print("=" * 60)
+
+    # Create model
+    model = BaramNuri(num_classes=5, pretrained=True)
+
+    # Parameter count
+    total_params = count_parameters(model)
+    print(f"\nTotal parameters: {total_params:,} ({total_params/1e6:.2f}M)")
+
+    # Test with dummy input
+    dummy_input = torch.randn(2, 3, 30, 224, 224)
+    print(f"\nInput shape: {dummy_input.shape}")
+
+    # Forward pass
+    model.eval()
+    with torch.no_grad():
+        output = model(dummy_input)
+    print(f"Output shape: {output.shape}")
+
+    # Single sample prediction test
+    single_input = torch.randn(1, 3, 30, 224, 224)
+    prediction = model.predict(single_input)
+    print(f"\nPrediction (Korean): {prediction['class_name']} ({prediction['confidence']:.2%})")
+
+    prediction_en = model.predict(single_input, return_english=True)
+    print(f"Prediction (English): {prediction_en['class_name']} ({prediction_en['confidence']:.2%})")
+
+    print("\nModel test passed!")

requirements.txt

@@ -0,0 +1,3 @@
+torch>=2.0.0
+torchvision>=0.15.0
+numpy>=1.21.0