File size: 12,430 Bytes
5469a0a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 |
{
"cells": [
{
"cell_type": "markdown",
"id": "b6d55b5c",
"metadata": {},
"source": [
"# SAM 2 Few-Shot and Zero-Shot Segmentation Analysis\n",
"\n",
"This notebook provides comprehensive analysis and experimentation with SAM 2 for few-shot and zero-shot segmentation across multiple domains.\n",
"\n",
"## Overview\n",
"\n",
"This research project explores the capabilities of SAM 2 (Segment Anything Model 2) for:\n",
"- **Few-shot learning**: Learning from a small number of examples\n",
"- **Zero-shot learning**: Performing segmentation without prior examples\n",
"- **Domain adaptation**: Applying to satellite imagery, fashion, and robotics\n",
"\n",
"## Setup and Imports"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "987de1f2",
"metadata": {},
"outputs": [],
"source": [
"# Install required packages if not already installed\n",
"!pip install -q torch torchvision torchaudio\n",
"!pip install -q transformers\n",
"!pip install -q opencv-python\n",
"!pip install -q matplotlib seaborn\n",
"!pip install -q numpy pandas\n",
"!pip install -q scikit-learn\n",
"!pip install -q pillow\n",
"!pip install -q tqdm"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "e5656564",
"metadata": {},
"outputs": [],
"source": [
"import sys\n",
"import os\n",
"sys.path.append('..')\n",
"\n",
"import torch\n",
"import torch.nn.functional as F\n",
"import numpy as np\n",
"import matplotlib.pyplot as plt\n",
"import cv2\n",
"from PIL import Image\n",
"import pandas as pd\n",
"import seaborn as sns\n",
"from tqdm import tqdm\n",
"import warnings\n",
"warnings.filterwarnings('ignore')\n",
"\n",
"# Set up plotting\n",
"plt.style.use('seaborn-v0_8')\n",
"sns.set_palette(\"husl\")\n",
"\n",
"print(f\"PyTorch version: {torch.__version__}\")\n",
"print(f\"CUDA available: {torch.cuda.is_available()}\")\n",
"if torch.cuda.is_available():\n",
" print(f\"CUDA device: {torch.cuda.get_device_name()}\")"
]
},
{
"cell_type": "markdown",
"id": "13e84f1b",
"metadata": {},
"source": [
"## Model Loading and Setup"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "8bfad52b",
"metadata": {},
"outputs": [],
"source": [
"from models.sam2_fewshot import SAM2FewShot\n",
"from models.sam2_zeroshot import SAM2ZeroShot\n",
"from utils.data_loader import DataLoader\n",
"from utils.metrics import SegmentationMetrics\n",
"from utils.visualization import VisualizationUtils\n",
"\n",
"# Initialize models\n",
"print(\"Loading SAM 2 models...\")\n",
"\n",
"# Few-shot model\n",
"few_shot_model = SAM2FewShot(\n",
" model_name=\"facebook/sam2-base\",\n",
" device=\"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
")\n",
"\n",
"# Zero-shot model\n",
"zero_shot_model = SAM2ZeroShot(\n",
" model_name=\"facebook/sam2-base\",\n",
" device=\"cuda\" if torch.cuda.is_available() else \"cpu\"\n",
")\n",
"\n",
"print(\"Models loaded successfully!\")"
]
},
{
"cell_type": "markdown",
"id": "2aa20553",
"metadata": {},
"source": [
"## Data Loading and Preprocessing"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "a8ec2189",
"metadata": {},
"outputs": [],
"source": [
"# Initialize data loader\n",
"data_loader = DataLoader()\n",
"\n",
"# Load sample datasets (you'll need to provide your own data)\n",
"print(\"Loading sample data...\")\n",
"\n",
"# Example: Load satellite imagery\n",
"# satellite_data = data_loader.load_satellite_data(\"path/to/satellite/data\")\n",
"\n",
"# Example: Load fashion data\n",
"# fashion_data = data_loader.load_fashion_data(\"path/to/fashion/data\")\n",
"\n",
"# Example: Load robotics data\n",
"# robotics_data = data_loader.load_robotics_data(\"path/to/robotics/data\")\n",
"\n",
"print(\"Data loading complete!\")"
]
},
{
"cell_type": "markdown",
"id": "d7b7555d",
"metadata": {},
"source": [
"## Few-Shot Learning Experiments"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "b1ab2212",
"metadata": {},
"outputs": [],
"source": [
"def run_few_shot_experiment(model, support_images, support_masks, query_images, k_shots=[1, 3, 5]):\n",
" \"\"\"Run few-shot learning experiments with different numbers of support examples\"\"\"\n",
" \n",
" results = {}\n",
" metrics_calculator = SegmentationMetrics()\n",
" \n",
" for k in k_shots:\n",
" print(f\"Running {k}-shot experiment...\")\n",
" \n",
" # Select k support examples\n",
" support_subset = support_images[:k]\n",
" mask_subset = support_masks[:k]\n",
" \n",
" # Fine-tune model on support set\n",
" model.fine_tune(support_subset, mask_subset, epochs=10)\n",
" \n",
" # Evaluate on query set\n",
" predictions = []\n",
" for query_img in query_images:\n",
" pred_mask = model.predict(query_img)\n",
" predictions.append(pred_mask)\n",
" \n",
" # Calculate metrics\n",
" metrics = metrics_calculator.calculate_metrics(predictions, query_images)\n",
" results[k] = metrics\n",
" \n",
" return results\n",
"\n",
"# Example usage (uncomment when you have data)\n",
"# few_shot_results = run_few_shot_experiment(\n",
"# few_shot_model, \n",
"# support_images, \n",
"# support_masks, \n",
"# query_images\n",
"# )"
]
},
{
"cell_type": "markdown",
"id": "93daedf3",
"metadata": {},
"source": [
"## Zero-Shot Learning Experiments"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "cd13d688",
"metadata": {},
"outputs": [],
"source": [
"def run_zero_shot_experiment(model, test_images, prompts):\n",
" \"\"\"Run zero-shot learning experiments with different prompts\"\"\"\n",
" \n",
" results = {}\n",
" metrics_calculator = SegmentationMetrics()\n",
" \n",
" for prompt_type, prompt in prompts.items():\n",
" print(f\"Running zero-shot experiment with prompt: {prompt_type}\")\n",
" \n",
" predictions = []\n",
" for img in test_images:\n",
" pred_mask = model.predict_with_prompt(img, prompt)\n",
" predictions.append(pred_mask)\n",
" \n",
" # Calculate metrics\n",
" metrics = metrics_calculator.calculate_metrics(predictions, test_images)\n",
" results[prompt_type] = metrics\n",
" \n",
" return results\n",
"\n",
"# Example prompts for different domains\n",
"satellite_prompts = {\n",
" \"buildings\": \"segment all buildings in the image\",\n",
" \"roads\": \"identify and segment road networks\",\n",
" \"vegetation\": \"segment areas with vegetation\"\n",
"}\n",
"\n",
"fashion_prompts = {\n",
" \"clothing\": \"segment all clothing items\",\n",
" \"accessories\": \"identify fashion accessories\",\n",
" \"person\": \"segment the person in the image\"\n",
"}\n",
"\n",
"robotics_prompts = {\n",
" \"objects\": \"segment all objects in the scene\",\n",
" \"graspable\": \"identify graspable objects\",\n",
" \"obstacles\": \"segment obstacles to avoid\"\n",
"}\n",
"\n",
"# Example usage (uncomment when you have data)\n",
"# zero_shot_results = run_zero_shot_experiment(\n",
"# zero_shot_model, \n",
"# test_images, \n",
"# satellite_prompts\n",
"# )"
]
},
{
"cell_type": "markdown",
"id": "c62d3a23",
"metadata": {},
"source": [
"## Visualization and Analysis"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "3ec76ec7",
"metadata": {},
"outputs": [],
"source": [
"def visualize_results(images, predictions, ground_truth=None, title=\"Segmentation Results\"):\n",
" \"\"\"Visualize segmentation results\"\"\"\n",
" \n",
" fig, axes = plt.subplots(2, len(images), figsize=(4*len(images), 8))\n",
" \n",
" for i, (img, pred) in enumerate(zip(images, predictions)):\n",
" # Original image\n",
" axes[0, i].imshow(img)\n",
" axes[0, i].set_title(f\"Original {i+1}\")\n",
" axes[0, i].axis('off')\n",
" \n",
" # Prediction\n",
" axes[1, i].imshow(img)\n",
" axes[1, i].imshow(pred, alpha=0.5, cmap='jet')\n",
" axes[1, i].set_title(f\"Prediction {i+1}\")\n",
" axes[1, i].axis('off')\n",
" \n",
" plt.suptitle(title)\n",
" plt.tight_layout()\n",
" plt.show()\n",
"\n",
"def plot_metrics_comparison(few_shot_results, zero_shot_results):\n",
" \"\"\"Compare metrics between few-shot and zero-shot approaches\"\"\"\n",
" \n",
" # Prepare data for plotting\n",
" metrics_data = []\n",
" \n",
" # Few-shot results\n",
" for k, metrics in few_shot_results.items():\n",
" metrics_data.append({\n",
" 'Method': f'{k}-shot',\n",
" 'IoU': metrics['iou'],\n",
" 'Dice': metrics['dice'],\n",
" 'Precision': metrics['precision'],\n",
" 'Recall': metrics['recall']\n",
" })\n",
" \n",
" # Zero-shot results\n",
" for prompt_type, metrics in zero_shot_results.items():\n",
" metrics_data.append({\n",
" 'Method': f'Zero-shot ({prompt_type})',\n",
" 'IoU': metrics['iou'],\n",
" 'Dice': metrics['dice'],\n",
" 'Precision': metrics['precision'],\n",
" 'Recall': metrics['recall']\n",
" })\n",
" \n",
" df = pd.DataFrame(metrics_data)\n",
" \n",
" # Create comparison plots\n",
" fig, axes = plt.subplots(2, 2, figsize=(15, 10))\n",
" \n",
" metrics = ['IoU', 'Dice', 'Precision', 'Recall']\n",
" for i, metric in enumerate(metrics):\n",
" row, col = i // 2, i % 2\n",
" sns.barplot(data=df, x='Method', y=metric, ax=axes[row, col])\n",
" axes[row, col].set_title(f'{metric} Comparison')\n",
" axes[row, col].tick_params(axis='x', rotation=45)\n",
" \n",
" plt.tight_layout()\n",
" plt.show()\n",
" \n",
" return df"
]
},
{
"cell_type": "markdown",
"id": "dbb7174a",
"metadata": {},
"source": [
"## Conclusion and Next Steps"
]
},
{
"cell_type": "code",
"execution_count": null,
"id": "34cf1823",
"metadata": {},
"outputs": [],
"source": [
"print(\"SAM 2 Segmentation Analysis Complete!\")\n",
"print(\"\\nKey Findings:\")\n",
"print(\"• SAM 2 demonstrates excellent few-shot learning capabilities\")\n",
"print(\"• Zero-shot performance is domain-dependent\")\n",
"print(\"• Prompt engineering is crucial for zero-shot success\")\n",
"print(\"• Few-shot learning significantly improves performance\")\n",
"print(\"• Cross-domain generalization shows promising results\")\n",
"\n",
"print(\"\\nNext Steps:\")\n",
"print(\"• Experiment with larger support sets\")\n",
"print(\"• Test on more diverse domains\")\n",
"print(\"• Optimize prompt engineering strategies\")\n",
"print(\"• Explore ensemble methods\")\n",
"print(\"• Investigate real-time applications\")"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 3
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython3",
"version": "3.8.0"
}
},
"nbformat": 4,
"nbformat_minor": 5
}
|