chore: Dockerfile, docs, deps, and comment tweaks

- Update Dockerfile, README, requirements, .gitignore
- Small comment/docstring adjustments in main, justify_thresholds, xgboost scripts

.gitignore

@@ -42,7 +42,10 @@ htmlcov/
 # Project specific
 focus_guard.db
 test_focus_guard.db
+<<<<<<< HEAD
+=======
 static/
+>>>>>>> feature/integration2.0
 __pycache__/
 docs/
 docs

Dockerfile

@@ -2,12 +2,20 @@ FROM python:3.10-slim
 
 RUN useradd -m -u 1000 user
 ENV HOME=/home/user PATH=/home/user/.local/bin:$PATH
+<<<<<<< HEAD
+=======
+
+>>>>>>> feature/integration2.0
 ENV PYTHONUNBUFFERED=1
 
 WORKDIR /app
 
 RUN apt-get update && apt-get install -y --no-install-recommends \
+<<<<<<< HEAD
     libglib2.0-0 libsm6 libxrender1 libxext6 libxcb1 libgl1 libgles2 libegl1 libgomp1 \
+=======
+    libglib2.0-0 libsm6 libxrender1 libxext6 libxcb1 libgl1 libgomp1 \
+>>>>>>> feature/integration2.0
     ffmpeg libavcodec-dev libavformat-dev libavutil-dev libswscale-dev \
     libavdevice-dev libopus-dev libvpx-dev libsrtp2-dev \
     build-essential nodejs npm git \
@@ -16,7 +24,9 @@ RUN apt-get update && apt-get install -y --no-install-recommends \
 RUN pip install --no-cache-dir torch torchvision --index-url https://download.pytorch.org/whl/cpu
 
 COPY requirements.txt ./
-RUN pip install --no-cache-dir -r requirements.txt
+<<<<<<< HEAD
+RUN pip install --no-cache-dir torch torchvision --index-url https://download.pytorch.org/whl/cpu \
+    && pip install --no-cache-dir -r requirements.txt
 
 COPY . .
 
@@ -26,10 +36,24 @@ RUN npm install && npm run build && mkdir -p /app/static && cp -R dist/* /app/st
 ENV FOCUSGUARD_CACHE_DIR=/app/.cache/focusguard
 RUN python -c "from models.face_mesh import _ensure_model; _ensure_model()"
 RUN python download_l2cs_weights.py || echo "[WARN] L2CS weights not downloaded — will run without gaze model"
+=======
+RUN pip install --no-cache-dir -r requirements.txt
+
+COPY . .
+
+RUN npm install && npm run build && mkdir -p /app/static && cp -R dist/* /app/static/
+
+ENV FOCUSGUARD_CACHE_DIR=/app/.cache/focusguard
+RUN python -c "from models.face_mesh import _ensure_model; _ensure_model()"
+>>>>>>> feature/integration2.0
 
 RUN mkdir -p /app/data && chown -R user:user /app
 
 USER user
 EXPOSE 7860
 
+<<<<<<< HEAD
 CMD ["bash", "start.sh"]
+=======
+CMD ["uvicorn", "main:app", "--host", "0.0.0.0", "--port", "7860", "--log-level", "info"]
+>>>>>>> feature/integration2.0

README.md

@@ -1,15 +1,21 @@
+<<<<<<< HEAD
 ---
 title: FocusGuard
 sdk: docker
 app_port: 7860
 ---
 
+=======
+>>>>>>> feature/integration2.0
 # FocusGuard
 
 Webcam-based focus detection: MediaPipe face mesh -> 17 features (EAR, gaze, head pose, PERCLOS, etc.) -> MLP or XGBoost for focused/unfocused. React + FastAPI app with WebSocket video.
 
+<<<<<<< HEAD
+=======
 **Repository:** Add your repo link here (e.g. `https://github.com/your-org/FocusGuard`).
 
+>>>>>>> feature/integration2.0
 ## Project layout
 
 ```
@@ -35,10 +41,13 @@ Webcam-based focus detection: MediaPipe face mesh -> 17 features (EAR, gaze, hea
 └── package.json
 ```
 
+<<<<<<< HEAD
+=======
 ## Config
 
 Hyperparameters and app settings live in `config/default.yaml` (learning rates, batch size, thresholds, L2CS weights, etc.). Override with env `FOCUSGUARD_CONFIG` pointing to another YAML.
 
+>>>>>>> feature/integration2.0
 ## Setup
 
 ```bash
@@ -86,6 +95,8 @@ python -m models.mlp.train
 python -m models.xgboost.train
 ```
 
+<<<<<<< HEAD
+=======
 ### ClearML experiment tracking
 
 All training and evaluation config (from `config/default.yaml`) is exposed as ClearML task parameters. Enable logging with `USE_CLEARML=1`; optionally run on a **remote GPU agent** instead of locally:
@@ -104,12 +115,16 @@ clearml-agent daemon --queue gpu
 
 Logged to ClearML: **parameters** (full flattened config), **scalars** (loss, accuracy, F1, ROC-AUC, per-class precision/recall/F1, dataset sizes and class counts), **artifacts** (best checkpoint, training log JSON), and **plots** (confusion matrix, ROC curves in evaluation).
 
+>>>>>>> feature/integration2.0
 ## Data
 
 9 participants, 144,793 samples, 10 features, binary labels. Collect with `python -m models.collect_features --name <name>`. Data lives in `data/collected_<name>/`.
 
+<<<<<<< HEAD
+=======
 **Train/val/test split:** All pooled training and evaluation use the same split for reproducibility. The test set is held out before any preprocessing; `StandardScaler` is fit on the training set only, then applied to val and test. Split ratios and random seed come from `config/default.yaml` (`data.split_ratios`, `mlp.seed`) via `data_preparation.prepare_dataset.get_default_split_config()`. MLP train, XGBoost train, eval_accuracy scripts, and benchmarks all use this single source so reported test accuracy is on the same held-out set.
 
+>>>>>>> feature/integration2.0
 ## Models
 
 | Model | What it uses | Best for |
@@ -127,6 +142,8 @@ Logged to ClearML: **parameters** (full flattened config), **scalars** (loss, ac
 | XGBoost (600 trees, depth 8) | 95.87% | 0.959 | 0.991 |
 | MLP (64->32) | 92.92% | 0.929 | 0.971 |
 
+<<<<<<< HEAD
+=======
 ## Model numbers (LOPO, 9 participants)
 
 | Model | LOPO AUC | Best threshold (Youden's J) | F1 @ best threshold | F1 @ 0.50 |
@@ -163,6 +180,7 @@ Latest quick feature-selection run (`python -m evaluation.feature_importance --q
 Top-5 XGBoost gain features: `s_face`, `ear_right`, `head_deviation`, `ear_avg`, `perclos`.
 For full leave-one-feature-out ablation, run `python -m evaluation.feature_importance` (slower).
 
+>>>>>>> feature/integration2.0
 ## L2CS Gaze Tracking
 
 L2CS-Net predicts where your eyes are looking, not just where your head is pointed. This catches the scenario where your head faces the screen but your eyes wander.

evaluation/justify_thresholds.py

@@ -80,7 +80,6 @@ def _plot_roc(fpr, tpr, auc, opt_thresh, opt_idx, title, path, clearml_title=Non
     ax.legend(loc="lower right")
     fig.tight_layout()
 
-    # Log to ClearML before closing the figure
     if _logger and clearml_title:
         _logger.report_matplotlib_figure(
             title=clearml_title, series="ROC", figure=fig, iteration=0
@@ -158,7 +157,6 @@ def analyse_model_thresholds(results):
         print(f"  {label}: AUC={auc:.4f}, optimal threshold={opt_t:.3f} "
               f"(F1={f1_opt:.4f}), F1@0.50={f1_50:.4f}")
 
-        # Log scalars to ClearML 
         if _logger:
             _logger.report_single_value(f"{label} Optimal Threshold", opt_t)
             _logger.report_single_value(f"{label} AUC", auc)
@@ -212,7 +210,6 @@ def run_geo_weight_search():
                 ha="center", va="bottom", fontsize=8)
     fig.tight_layout()
 
-    # Log to ClearML before closing
     if _logger:
         _logger.report_matplotlib_figure(
             title="Geo Weight Search", series="F1 vs Alpha", figure=fig, iteration=0
@@ -226,7 +223,6 @@ def run_geo_weight_search():
     print(f"  Best alpha (face weight) = {best_alpha:.1f}, "
           f"mean LOPO F1 = {mean_f1[best_alpha]:.4f}")
 
-    # Log scalars to ClearML 
     if _logger:
         _logger.report_single_value("Geo Best Alpha", best_alpha)
         for i, a in enumerate(sorted(alphas)):
@@ -302,7 +298,6 @@ def run_hybrid_weight_search(lopo_results):
                 ha="center", va="bottom", fontsize=8)
     fig.tight_layout()
 
-    # Log to ClearML before closing
     if _logger:
         _logger.report_matplotlib_figure(
             title="Hybrid Weight Search", series="F1 vs w_mlp", figure=fig, iteration=0
@@ -315,7 +310,6 @@ def run_hybrid_weight_search(lopo_results):
 
     print(f"  Best w_mlp = {best_w:.1f}, mean LOPO F1 = {mean_f1[best_w]:.4f}")
 
-    # Log scalars to ClearML 
     if _logger:
         _logger.report_single_value("Hybrid Best w_mlp", best_w)
         for i, w in enumerate(sorted(w_mlps)):
@@ -366,7 +360,6 @@ def plot_distributions():
     ax.legend(fontsize=8)
     fig_ear.tight_layout()
 
-    # Log to ClearML before closing
     if _logger:
         _logger.report_matplotlib_figure(
             title="EAR Distribution", series="by class", figure=fig_ear, iteration=0
@@ -388,7 +381,6 @@ def plot_distributions():
     ax.legend(fontsize=8)
     fig_mar.tight_layout()
 
-    # Log to ClearML before closing
     if _logger:
         _logger.report_matplotlib_figure(
             title="MAR Distribution", series="by class", figure=fig_mar, iteration=0
@@ -550,7 +542,6 @@ def main():
 
     write_report(model_stats, geo_f1, best_alpha, hybrid_f1, best_w, dist_stats)
 
-    # Close ClearML task
     if _task:
         from config.clearml_enrich import task_done_summary
 

main.py

@@ -110,7 +110,6 @@ async def lifespan(app):
 
 app = FastAPI(title="Focus Guard API", lifespan=lifespan)
 
-# Add CORS middleware
 app.add_middleware(
     CORSMiddleware,
     allow_origins=["*"],
@@ -119,7 +118,6 @@ app.add_middleware(
     allow_headers=["*"],
 )
 
-# Global variables
 pcs = set()
 _cached_model_name = get("app.default_model") or "mlp"
 _l2cs_boost_enabled = False
@@ -198,7 +196,6 @@ class VideoTransformTrack(VideoStreamTrack):
                     "model": model_name,
                 }
 
-                # Draw face mesh + HUD on the video frame
                 h_f, w_f = img.shape[:2]
                 lm = out.get("landmarks")
                 eye_gaze_enabled = _l2cs_boost_enabled or model_name == "l2cs"
@@ -302,7 +299,6 @@ _BOOST_VETO = 0.38  # L2CS below this -> forced not-focused
 
 
 def _process_frame_with_l2cs_boost(base_pipeline, frame, base_model_name):
-    # run base model
     with _pipeline_locks[base_model_name]:
         base_out = base_pipeline.process_frame(frame)
 
@@ -311,7 +307,6 @@ def _process_frame_with_l2cs_boost(base_pipeline, frame, base_model_name):
         base_out["boost_active"] = False
         return base_out
 
-    # run L2CS
     with _pipeline_locks["l2cs"]:
         l2cs_out = l2cs_pipe.process_frame(frame)
 
@@ -804,7 +799,7 @@ async def import_sessions(sessions: List[dict]):
 async def clear_history():
     try:
         async with aiosqlite.connect(db_path) as db:
-            # Delete events first (foreign key good practice)
+            # events reference sessions via FK
             await db.execute("DELETE FROM focus_events")
             await db.execute("DELETE FROM focus_sessions")
             await db.commit()

models/xgboost/add_accuracy.py

@@ -16,7 +16,6 @@ X_val,   y_val   = splits["X_val"],   splits["y_val"]
 csv_path = 'models/xgboost/sweep_results_all_40.csv'
 df = pd.read_csv(csv_path)
 
-# We will calculate accuracy for each row
 accuracies = []
 
 print(f"Re-evaluating {len(df)} configurations for accuracy. This will take a few minutes...")
@@ -34,7 +33,6 @@ for idx, row in df.iterrows():
         "eval_metric": "logloss"
     }
     
-    # Train the exact same model quickly
     model = XGBClassifier(**params)
     model.fit(X_train, y_train)
     
@@ -46,12 +44,10 @@ for idx, row in df.iterrows():
     if (idx + 1) % 5 == 0:
         print(f"Processed {idx + 1}/{len(df)} trials...")
 
-# Add accuracy column and save back to CSV
 df.insert(2, 'val_accuracy', accuracies)
 df.to_csv(csv_path, index=False)
 
 print(f"\nDone! Updated {csv_path} with 'val_accuracy'.")
-# Display the top 5 by accuracy now just to see
 top5_acc = df.nlargest(5, 'val_accuracy')[['task_id', 'val_accuracy', 'val_f1', 'val_loss']]
 print("\nTop 5 Trials by Accuracy:")
 print(top5_acc.to_string(index=False))

models/xgboost/sweep_local.py

@@ -25,6 +25,7 @@ DATA_SPLITS, SEED = get_default_split_config()
 
 # ── Search Space ──────────────────────────────────────────────────────────────
 def objective(trial):
+    # 1. Sample hyperparameters
     # 1. Sample hyperparameters
     params = {
         "n_estimators": trial.suggest_categorical("n_estimators", [100, 200, 400, 600]),

requirements.txt

@@ -16,11 +16,17 @@ httpx>=0.27.0
 aiosqlite>=0.19.0
 psutil>=5.9.0
 pydantic>=2.0.0
+<<<<<<< HEAD
+=======
 PyYAML>=6.0
+>>>>>>> feature/integration2.0
 xgboost>=2.0.0
 clearml>=2.0.2
 pytest>=9.0.0
 pytest-cov>=5.0.0
 face_detection @ git+https://github.com/elliottzheng/face-detection
 gdown>=5.0.0
+<<<<<<< HEAD
 huggingface_hub
+=======
+>>>>>>> feature/integration2.0