|
| 1 | +<!DOCTYPE html> |
| 2 | +<html lang="en"> |
| 3 | +<head> |
| 4 | + <meta charset="UTF-8"> |
| 5 | + <meta name="viewport" content="width=device-width, initial-scale=1.0"> |
| 6 | + <title>Gemini API Image Processor with Bounding Box Visualization</title> |
| 7 | + <script type="module"> |
| 8 | + import { GoogleGenerativeAI } from "https://esm.run/@google/generative-ai"; |
| 9 | + import { marked } from "https://esm.run/marked"; |
| 10 | + |
| 11 | + function getApiKey() { |
| 12 | + let apiKey = localStorage.getItem("GEMINI_API_KEY"); |
| 13 | + if (!apiKey) { |
| 14 | + apiKey = prompt("Please enter your Gemini API key:"); |
| 15 | + if (apiKey) { |
| 16 | + localStorage.setItem("GEMINI_API_KEY", apiKey); |
| 17 | + } |
| 18 | + } |
| 19 | + return apiKey; |
| 20 | + } |
| 21 | + |
| 22 | + async function getGenerativeModel(params) { |
| 23 | + const API_KEY = getApiKey(); |
| 24 | + const genAI = new GoogleGenerativeAI(API_KEY); |
| 25 | + return genAI.getGenerativeModel(params); |
| 26 | + } |
| 27 | + |
| 28 | + async function fileToGenerativePart(file) { |
| 29 | + return new Promise((resolve) => { |
| 30 | + const reader = new FileReader(); |
| 31 | + reader.onloadend = () => resolve({ |
| 32 | + inlineData: { |
| 33 | + data: reader.result.split(",")[1], |
| 34 | + mimeType: file.type |
| 35 | + } |
| 36 | + }); |
| 37 | + reader.readAsDataURL(file); |
| 38 | + }); |
| 39 | + } |
| 40 | + |
| 41 | + async function processImageAndPrompt() { |
| 42 | + const fileInput = document.getElementById('imageInput'); |
| 43 | + const promptInput = document.getElementById('promptInput'); |
| 44 | + const resultDiv = document.getElementById('result'); |
| 45 | + |
| 46 | + if (!fileInput.files[0] || !promptInput.value) { |
| 47 | + alert('Please select an image and enter a prompt.'); |
| 48 | + return; |
| 49 | + } |
| 50 | + |
| 51 | + resultDiv.innerHTML = 'Processing...'; |
| 52 | + |
| 53 | + try { |
| 54 | + const model = await getGenerativeModel({ model: "gemini-1.5-pro" }); |
| 55 | + const imagePart = await fileToGenerativePart(fileInput.files[0]); |
| 56 | + |
| 57 | + const result = await model.generateContent([promptInput.value, imagePart]); |
| 58 | + const response = await result.response; |
| 59 | + const text = response.text(); |
| 60 | + |
| 61 | + resultDiv.innerHTML = marked.parse(text); |
| 62 | + |
| 63 | + // Extract coordinates from the response |
| 64 | + const coordinates = extractCoordinates(text); |
| 65 | + if (coordinates.length > 0) { |
| 66 | + displayImageWithBoundingBoxes(fileInput.files[0], coordinates); |
| 67 | + } |
| 68 | + } catch (error) { |
| 69 | + resultDiv.innerHTML = `Error: ${error.message}`; |
| 70 | + } |
| 71 | + } |
| 72 | + |
| 73 | + function extractCoordinates(text) { |
| 74 | + const regex = /\[\s*\d+\s*,\s*\d+\s*,\s*\d+\s*,\s*\d+\s*\]/g; |
| 75 | + const matches = text.match(regex) || []; |
| 76 | + return matches.map(JSON.parse); |
| 77 | + } |
| 78 | + |
| 79 | + function displayImageWithBoundingBoxes(file, coordinates) { |
| 80 | + const reader = new FileReader(); |
| 81 | + reader.onload = function(event) { |
| 82 | + const image = new Image(); |
| 83 | + image.onload = function() { |
| 84 | + const canvas = document.getElementById('canvas'); |
| 85 | + canvas.width = image.width; |
| 86 | + canvas.height = image.height; |
| 87 | + const ctx = canvas.getContext('2d'); |
| 88 | + ctx.drawImage(image, 0, 0); |
| 89 | + |
| 90 | + const colors = ['#FF0000', '#00FF00', '#0000FF', '#FFFF00', '#FF00FF', '#00FFFF']; |
| 91 | + coordinates.forEach((box, index) => { |
| 92 | + const [ymin, xmin, ymax, xmax] = box.map(coord => (1000 - coord) / 1000); |
| 93 | + const width = (xmax - xmin) * image.width; |
| 94 | + const height = (ymax - ymin) * image.height; |
| 95 | + |
| 96 | + ctx.strokeStyle = colors[index % colors.length]; |
| 97 | + ctx.lineWidth = 2; |
| 98 | + ctx.strokeRect(xmin * image.width, ymin * image.height, width, height); |
| 99 | + }); |
| 100 | + }; |
| 101 | + image.src = event.target.result; |
| 102 | + }; |
| 103 | + reader.readAsDataURL(file); |
| 104 | + } |
| 105 | + |
| 106 | + // Attach event listener to the submit button |
| 107 | + document.getElementById('submitBtn').addEventListener('click', processImageAndPrompt); |
| 108 | + </script> |
| 109 | + <style> |
| 110 | + body { |
| 111 | + font-family: Arial, sans-serif; |
| 112 | + max-width: 800px; |
| 113 | + margin: 0 auto; |
| 114 | + padding: 20px; |
| 115 | + } |
| 116 | + textarea { |
| 117 | + width: 100%; |
| 118 | + height: 100px; |
| 119 | + } |
| 120 | + #result, #imageContainer { |
| 121 | + margin-top: 20px; |
| 122 | + border: 1px solid #ccc; |
| 123 | + padding: 10px; |
| 124 | + } |
| 125 | + #canvas { |
| 126 | + max-width: 100%; |
| 127 | + height: auto; |
| 128 | + } |
| 129 | + </style> |
| 130 | +</head> |
| 131 | +<body> |
| 132 | + <h1>Gemini API Image Processor with Bounding Box Visualization</h1> |
| 133 | + <input type="file" id="imageInput" accept="image/*"> |
| 134 | + <textarea id="promptInput">Return bounding boxes as [ymin, xmin, ymax, xmax] |
| 135 | +</textarea> |
| 136 | + <button id="submitBtn">Process</button> |
| 137 | + <div id="result"></div> |
| 138 | + <div id="imageContainer"> |
| 139 | + <canvas id="canvas"></canvas> |
| 140 | + </div> |
| 141 | +</body> |
| 142 | +</html> |
0 commit comments