Spaces:
Paused
Paused
File size: 3,765 Bytes
95e7952 | 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 | {
"nbformat": 4,
"nbformat_minor": 0,
"metadata": {
"colab": {
"provenance": []
},
"kernelspec": {
"name": "python3",
"display_name": "Python 3"
},
"language_info": {
"name": "python"
}
},
"cells": [
{
"cell_type": "markdown",
"source": [
"# DeepScan: Production-Grade Side-Scan Sonar AI Training\n",
"**Smart India Hackathon 2026 | Problem Statement 26057**\n",
"\n",
"This notebook uses **Heavy Acoustic Augmentation** and Domain Adaptation to train a highly robust YOLOv8 model for AUV Edge deployment, even on small datasets."
],
"metadata": {
"id": "header"
}
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"id": "install"
},
"outputs": [],
"source": [
"!pip install ultralytics roboflow opencv-python\n",
"import torch\n",
"print(\"GPU Available:\", torch.cuda.is_available())"
]
},
{
"cell_type": "markdown",
"source": [
"### 1. Download Dataset via Roboflow\n",
"Replace the API key with your own from Roboflow Universe."
],
"metadata": {
"id": "dataset_md"
}
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"id": "dataset"
},
"outputs": [],
"source": [
"from roboflow import Roboflow\n",
"rf = Roboflow(api_key=\"YOUR_ROBOFLOW_API_KEY_HERE\")\n",
"project = rf.workspace(\"underwater-object-detection\").project(\"side-scan-sonar-debris\")\n",
"version = project.version(1)\n",
"dataset = version.download(\"yolov8\")"
]
},
{
"cell_type": "markdown",
"source": [
"### 2. Train with Heavy Acoustic Augmentation\n",
"Standard YOLO crushes sonar data. We use massive augmentations to simulate murky water, varying sonar frequencies, and AUV roll/pitch."
],
"metadata": {
"id": "train_md"
}
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"id": "train"
},
"outputs": [],
"source": [
"from ultralytics import YOLO\n",
"\n",
"# Load a pretrained YOLOv8 small model (optimal for Nvidia Jetson Edge deployment)\n",
"model = YOLO('yolov8s.pt')\n",
"\n",
"# Train the model with acoustic physics parameters\n",
"results = model.train(\n",
" data=f\"{dataset.location}/data.yaml\",\n",
" epochs=100,\n",
" imgsz=640,\n",
" batch=16,\n",
" project=\"models\",\n",
" name=\"sss_detector_v1\",\n",
" \n",
" # --- HEAVY ACOUSTIC AUGMENTATION ---\n",
" mosaic=1.0, # Stitches images to teach scale variation\n",
" mixup=0.2, # Blends anomalies into different backgrounds\n",
" hsv_h=0.015, # Simulates changing sonar ping frequencies\n",
" hsv_s=0.7, \n",
" hsv_v=0.4, # Simulates attenuation (darkness) in deep water\n",
" degrees=10.0, # Simulates AUV rolling due to currents\n",
" translate=0.1, # Simulates AUV shifting\n",
" flipud=0.5, # Common in bidirectional sonar scans\n",
" fliplr=0.5\n",
")"
]
},
{
"cell_type": "markdown",
"source": [
"### 3. Export for Deployment\n",
"Once training is complete, download `models/sss_detector_v1/weights/best.pt` and place it in your local React/FastAPI project!"
],
"metadata": {
"id": "export_md"
}
}
]
}
|