separated notebooks for clarity

notebooks/demo.ipynb

@@ -1,248 +0,0 @@
-{
- "cells": [
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "# S2F Model Evaluation\n",
-    "\n",
-    "This notebook shows how to evaluate a trained Shape2Force (S2F) model on your dataset.\n",
-    "\n",
-    "**Metrics computed:**\n",
-    "- **MSE** – Mean Squared Error\n",
-    "- **MS-SSIM** – Multi-Scale Structural Similarity\n",
-    "- **Pixel Correlation** – Pearson correlation between predicted and ground-truth heatmaps\n",
-    "- **Relative Magnitude Error** – WFM-style weighted relative error\n",
-    "- **Force Sum/Mean Correlation** – Correlation of total force per sample\n",
-    "\n",
-    "Run the cells below after adjusting paths and settings."
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "%load_ext autoreload\n",
-    "%autoreload 2\n",
-    "\n",
-    "import warnings\n",
-    "warnings.filterwarnings('ignore')\n",
-    "import sys\n",
-    "import os\n",
-    "import cv2\n",
-    "cv2.utils.logging.setLogLevel(cv2.utils.logging.LOG_LEVEL_ERROR)\n",
-    "\n",
-    "cwd = os.getcwd()\n",
-    "S2F_ROOT = cwd if os.path.exists(os.path.join(cwd, 'models')) else os.path.dirname(cwd)\n",
-    "sys.path.insert(0, S2F_ROOT)\n",
-    "\n",
-    "import torch\n",
-    "import matplotlib.pyplot as plt\n",
-    "\n",
-    "from data.cell_dataset import prepare_data, load_folder_data\n",
-    "from models.s2f_model import create_s2f_model, compute_settings_normalization\n",
-    "from utils.metrics import (\n",
-    "    evaluate_metrics_on_dataset,\n",
-    "    print_metrics_report,\n",
-    "    gen_prediction_plots,\n",
-    "    plot_predictions,\n",
-    ")\n",
-    "\n",
-    "print(f\"S2F root: {S2F_ROOT}\")"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "## Configuration"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# --- Adjust these paths ---\n",
-    "USE_SINGLE_CELL = True  # True = single-cell (with substrate), False = spheroid\n",
-    "DATA_FOLDER = os.path.join(S2F_ROOT, 'sample')  # or path to your dataset\n",
-    "CHECKPOINT_PATH = os.path.join(S2F_ROOT, 'ckp', 'best_checkpoint.pth')\n",
-    "\n",
-    "IMAGE_SIZE = 1024\n",
-    "BATCH_SIZE = 2\n",
-    "THRESHOLD = 0.0  # Threshold for heatmap metrics (0 = no threshold)\n",
-    "DEVICE = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
-    "SUBSTRATE = 'fibroblasts_PDMS'  # For single-cell mode"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "## Load Data"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# Option A: Load from folder with train/test subfolders\n",
-    "train_folder = os.path.join(DATA_FOLDER, 'train')\n",
-    "test_folder = os.path.join(DATA_FOLDER, 'test')\n",
-    "\n",
-    "if os.path.exists(train_folder) and os.path.exists(test_folder):\n",
-    "    train_loader, val_loader = prepare_data(\n",
-    "        DATA_FOLDER,\n",
-    "        batch_size=BATCH_SIZE,\n",
-    "        target_size=(IMAGE_SIZE, IMAGE_SIZE),\n",
-    "        use_augmentations=False,\n",
-    "        train_test_sep_folder=True,\n",
-    "        return_metadata=USE_SINGLE_CELL,\n",
-    "        substrate=SUBSTRATE if USE_SINGLE_CELL else None,\n",
-    "    )\n",
-    "    print(f\"Loaded train: {len(train_loader.dataset)} samples, val: {len(val_loader.dataset)} samples\")\n",
-    "else:\n",
-    "    # Option B: Load from a single folder (e.g. test only)\n",
-    "    val_loader = load_folder_data(\n",
-    "        DATA_FOLDER,\n",
-    "        substrate=SUBSTRATE if USE_SINGLE_CELL else None,\n",
-    "        img_size=IMAGE_SIZE,\n",
-    "        batch_size=BATCH_SIZE,\n",
-    "        return_metadata=USE_SINGLE_CELL,\n",
-    "    )\n",
-    "    train_loader = None\n",
-    "    print(f\"Loaded {len(val_loader.dataset)} samples from {DATA_FOLDER}\")"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "## Load Model"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "in_channels = 3 if USE_SINGLE_CELL else 1\n",
-    "generator, _ = create_s2f_model(in_channels=in_channels)\n",
-    "checkpoint = torch.load(CHECKPOINT_PATH, map_location='cpu', weights_only=False)\n",
-    "generator.load_state_dict(checkpoint.get('generator_state_dict', checkpoint), strict=True)\n",
-    "generator = generator.to(DEVICE)\n",
-    "generator.eval()\n",
-    "\n",
-    "print(f\"Loaded checkpoint from {CHECKPOINT_PATH}\")\n",
-    "if 'epoch' in checkpoint:\n",
-    "    print(f\"  Epoch: {checkpoint['epoch']}\")"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "## Run Evaluation"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "config_path = os.path.join(S2F_ROOT, 'config', 'substrate_settings.json')\n",
-    "normalization_params = compute_settings_normalization(config_path=config_path) if USE_SINGLE_CELL else None\n",
-    "\n",
-    "val_results = evaluate_metrics_on_dataset(\n",
-    "    generator,\n",
-    "    val_loader,\n",
-    "    device=DEVICE,\n",
-    "    description=\"Evaluating\",\n",
-    "    save_predictions=True,\n",
-    "    threshold=THRESHOLD,\n",
-    "    use_settings=USE_SINGLE_CELL,\n",
-    "    normalization_params=normalization_params,\n",
-    "    config_path=config_path,\n",
-    "    substrate_override=SUBSTRATE,\n",
-    ")\n",
-    "\n",
-    "report = {'validation': val_results}\n",
-    "if train_loader is not None:\n",
-    "    train_results = evaluate_metrics_on_dataset(\n",
-    "        generator, train_loader, device=DEVICE, description=\"Training\",\n",
-    "        threshold=THRESHOLD, use_settings=USE_SINGLE_CELL,\n",
-    "        normalization_params=normalization_params, config_path=config_path,\n",
-    "        substrate_override=SUBSTRATE,\n",
-    "    )\n",
-    "    report = {'train': train_results, 'validation': val_results}\n",
-    "\n",
-    "print_metrics_report(report, threshold=THRESHOLD)"
-   ]
-  },
-  {
-   "cell_type": "markdown",
-   "metadata": {},
-   "source": [
-    "## Visualize Predictions"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# Quick preview: plot first few samples\n",
-    "plot_predictions(\n",
-    "    val_loader,\n",
-    "    generator,\n",
-    "    n_samples=3,\n",
-    "    device=DEVICE,\n",
-    "    use_settings=USE_SINGLE_CELL,\n",
-    "    normalization_params=normalization_params,\n",
-    "    config_path=config_path,\n",
-    "    substrate_override=SUBSTRATE,\n",
-    ")"
-   ]
-  },
-  {
-   "cell_type": "code",
-   "execution_count": null,
-   "metadata": {},
-   "outputs": [],
-   "source": [
-    "# Save individual prediction plots (sorted by MSE, best first)\n",
-    "plot_output_dir = os.path.join(S2F_ROOT, 'outputs', 'evaluation_plots')\n",
-    "if 'individual_predictions' in val_results:\n",
-    "    gen_prediction_plots(\n",
-    "        val_results['individual_predictions'],\n",
-    "        save_dir=plot_output_dir,\n",
-    "        sort_by='mse',\n",
-    "        sort_order='asc',\n",
-    "        threshold=THRESHOLD,\n",
-    "    )\n",
-    "    print(f\"Saved plots to {plot_output_dir}\")"
-   ]
-  }
- ],
- "metadata": {
-  "kernelspec": {
-   "display_name": "Python 3",
-   "language": "python",
-   "name": "python3"
-  },
-  "language_info": {
-   "name": "python",
-   "version": "3.10.0"
-  }
- },
- "nbformat": 4,
- "nbformat_minor": 4
-}