Spaces:
Running
Running
Update src/components/MultiSourceCaptioningView.tsx
Browse files
src/components/MultiSourceCaptioningView.tsx
CHANGED
|
@@ -1,520 +1,520 @@
|
|
| 1 |
-
import { useState, useRef, useEffect } from "react";
|
| 2 |
-
import { useVLMContext } from "../context/useVLMContext";
|
| 3 |
-
import { extractJsonFromMarkdown, drawBoundingBoxesOnCanvas } from "./BoxAnnotator";
|
| 4 |
-
|
| 5 |
-
const MODES = ["Webcam", "URL", "File"] as const;
|
| 6 |
-
type Mode = typeof MODES[number];
|
| 7 |
-
|
| 8 |
-
const EXAMPLE_VIDEO_URL = "/
|
| 9 |
-
const EXAMPLE_PROMPT = "Detect all people in the image. For each person, output a JSON array of objects with fields: 'label' (string) and 'bbox_2d' ([x1, y1, x2, y2]) where coordinates are in pixel values. Example: [{\"label\": \"person\", \"bbox_2d\": [100, 50, 200, 300]}]";
|
| 10 |
-
|
| 11 |
-
function parseFlatBoxArray(arr: any[]): { label: string, bbox_2d: number[] }[] {
|
| 12 |
-
if (typeof arr[0] === "string" && Array.isArray(arr[1])) {
|
| 13 |
-
const label = arr[0];
|
| 14 |
-
return arr.slice(1).map(bbox => ({ label, bbox_2d: bbox }));
|
| 15 |
-
}
|
| 16 |
-
return [];
|
| 17 |
-
}
|
| 18 |
-
|
| 19 |
-
function normalizeBoxes(raw: any): { label: string, bbox_2d: number[] }[] {
|
| 20 |
-
if (!raw) return [];
|
| 21 |
-
let boxes = [];
|
| 22 |
-
if (typeof raw === "object" && raw !== null && Array.isArray(raw.image)) {
|
| 23 |
-
boxes = raw.image;
|
| 24 |
-
} else if (Array.isArray(raw)) {
|
| 25 |
-
boxes = raw;
|
| 26 |
-
} else if (typeof raw === "object" && raw !== null) {
|
| 27 |
-
boxes = [raw];
|
| 28 |
-
}
|
| 29 |
-
return boxes
|
| 30 |
-
.map((obj: any) => {
|
| 31 |
-
if (!obj || !obj.bbox_2d) return null;
|
| 32 |
-
let bbox = obj.bbox_2d;
|
| 33 |
-
// If bbox_2d is [[x1, y1], [x2, y2]], convert to [x1, y1, x2, y2]
|
| 34 |
-
if (
|
| 35 |
-
Array.isArray(bbox) &&
|
| 36 |
-
bbox.length === 2 &&
|
| 37 |
-
Array.isArray(bbox[0]) &&
|
| 38 |
-
Array.isArray(bbox[1]) &&
|
| 39 |
-
bbox[0].length === 2 &&
|
| 40 |
-
bbox[1].length === 2
|
| 41 |
-
) {
|
| 42 |
-
bbox = [bbox[0][0], bbox[0][1], bbox[1][0], bbox[1][1]];
|
| 43 |
-
}
|
| 44 |
-
// If bbox_2d is [x1, y1, x2, y2], use as-is
|
| 45 |
-
if (
|
| 46 |
-
Array.isArray(bbox) &&
|
| 47 |
-
bbox.length === 4 &&
|
| 48 |
-
bbox.every((v: any) => typeof v === "number")
|
| 49 |
-
) {
|
| 50 |
-
return { ...obj, bbox_2d: bbox };
|
| 51 |
-
}
|
| 52 |
-
// Otherwise, skip
|
| 53 |
-
return null;
|
| 54 |
-
})
|
| 55 |
-
.filter((obj: any) => obj);
|
| 56 |
-
}
|
| 57 |
-
|
| 58 |
-
function isImageFile(file: File) {
|
| 59 |
-
return file.type.startsWith("image/");
|
| 60 |
-
}
|
| 61 |
-
function isVideoFile(file: File) {
|
| 62 |
-
return file.type.startsWith("video/");
|
| 63 |
-
}
|
| 64 |
-
|
| 65 |
-
export default function MultiSourceCaptioningView() {
|
| 66 |
-
const [mode, setMode] = useState<Mode>("File");
|
| 67 |
-
const [videoUrl, setVideoUrl] = useState<string>(EXAMPLE_VIDEO_URL);
|
| 68 |
-
const [inputUrl, setInputUrl] = useState<string>(EXAMPLE_VIDEO_URL);
|
| 69 |
-
const [prompt, setPrompt] = useState<string>(EXAMPLE_PROMPT);
|
| 70 |
-
const [processing, setProcessing] = useState(false);
|
| 71 |
-
const [error, setError] = useState<string | null>(null);
|
| 72 |
-
const [webcamActive, setWebcamActive] = useState(false);
|
| 73 |
-
const [uploadedFile, setUploadedFile] = useState<File | null>(null);
|
| 74 |
-
const [uploadedUrl, setUploadedUrl] = useState<string>("");
|
| 75 |
-
const [videoProcessing, setVideoProcessing] = useState(false);
|
| 76 |
-
const [imageProcessed, setImageProcessed] = useState(false);
|
| 77 |
-
const [exampleProcessing, setExampleProcessing] = useState(false);
|
| 78 |
-
const [urlProcessing, setUrlProcessing] = useState(false);
|
| 79 |
-
const [debugOutput, setDebugOutput] = useState<string>("");
|
| 80 |
-
const [canvasDims, setCanvasDims] = useState<{w:number,h:number}|null>(null);
|
| 81 |
-
const [videoDims, setVideoDims] = useState<{w:number,h:number}|null>(null);
|
| 82 |
-
const [inferenceStatus, setInferenceStatus] = useState<string>("");
|
| 83 |
-
|
| 84 |
-
const videoRef = useRef<HTMLVideoElement | null>(null);
|
| 85 |
-
const canvasRef = useRef<HTMLCanvasElement | null>(null);
|
| 86 |
-
const imageRef = useRef<HTMLImageElement | null>(null);
|
| 87 |
-
const webcamStreamRef = useRef<MediaStream | null>(null);
|
| 88 |
-
const { isLoaded, isLoading, error: modelError, runInference } = useVLMContext();
|
| 89 |
-
|
| 90 |
-
const processVideoFrame = async () => {
|
| 91 |
-
if (!videoRef.current || !canvasRef.current) return;
|
| 92 |
-
const video = videoRef.current;
|
| 93 |
-
const canvas = canvasRef.current;
|
| 94 |
-
if (video.paused || video.ended || video.videoWidth === 0) return;
|
| 95 |
-
canvas.width = video.videoWidth;
|
| 96 |
-
canvas.height = video.videoHeight;
|
| 97 |
-
const ctx = canvas.getContext("2d");
|
| 98 |
-
if (!ctx) return;
|
| 99 |
-
ctx.drawImage(video, 0, 0, canvas.width, canvas.height);
|
| 100 |
-
await runInference(video, prompt, (output: string) => {
|
| 101 |
-
setDebugOutput(output); // <-- Ensure Raw Model Output is updated
|
| 102 |
-
let boxes = extractJsonFromMarkdown(output) || [];
|
| 103 |
-
if (boxes.length === 0 && Array.isArray(output)) {
|
| 104 |
-
boxes = parseFlatBoxArray(output);
|
| 105 |
-
}
|
| 106 |
-
boxes = normalizeBoxes(boxes);
|
| 107 |
-
console.log("Model output:", output);
|
| 108 |
-
console.log("Boxes after normalization:", boxes);
|
| 109 |
-
console.log("Canvas size:", canvas.width, canvas.height);
|
| 110 |
-
if (boxes.length > 0) {
|
| 111 |
-
const [x1, y1, x2, y2] = boxes[0].bbox_2d;
|
| 112 |
-
console.log("First box coords:", x1, y1, x2, y2);
|
| 113 |
-
}
|
| 114 |
-
if (boxes.length === 0) setInferenceStatus("No boxes detected or model output invalid.");
|
| 115 |
-
if (Array.isArray(boxes) && boxes.length > 0) {
|
| 116 |
-
const scaleX = canvas.width / video.videoWidth;
|
| 117 |
-
const scaleY = canvas.height / video.videoHeight;
|
| 118 |
-
ctx.clearRect(0, 0, canvas.width, canvas.height); // Clear canvas before drawing boxes
|
| 119 |
-
drawBoundingBoxesOnCanvas(ctx, boxes, { color: "#FF00FF", lineWidth: 4, font: "20px Arial", scaleX, scaleY }); // Use visible color and thick line
|
| 120 |
-
}
|
| 121 |
-
});
|
| 122 |
-
};
|
| 123 |
-
|
| 124 |
-
const handleFileChange = (e: React.ChangeEvent<HTMLInputElement>) => {
|
| 125 |
-
const file = e.target.files?.[0] || null;
|
| 126 |
-
setUploadedFile(file);
|
| 127 |
-
setUploadedUrl(file ? URL.createObjectURL(file) : "");
|
| 128 |
-
setError(null);
|
| 129 |
-
setImageProcessed(false);
|
| 130 |
-
setVideoProcessing(false);
|
| 131 |
-
setExampleProcessing(false);
|
| 132 |
-
};
|
| 133 |
-
|
| 134 |
-
// Webcam setup and teardown (unchanged)
|
| 135 |
-
useEffect(() => {
|
| 136 |
-
if (mode !== "Webcam") {
|
| 137 |
-
if (webcamStreamRef.current) {
|
| 138 |
-
webcamStreamRef.current.getTracks().forEach((track: MediaStreamTrack) => track.stop());
|
| 139 |
-
webcamStreamRef.current = null;
|
| 140 |
-
}
|
| 141 |
-
setWebcamActive(false);
|
| 142 |
-
return;
|
| 143 |
-
}
|
| 144 |
-
const setupWebcam = async () => {
|
| 145 |
-
try {
|
| 146 |
-
setError(null);
|
| 147 |
-
const stream = await navigator.mediaDevices.getUserMedia({ video: true });
|
| 148 |
-
webcamStreamRef.current = stream;
|
| 149 |
-
if (videoRef.current) {
|
| 150 |
-
videoRef.current.srcObject = stream;
|
| 151 |
-
setWebcamActive(true);
|
| 152 |
-
}
|
| 153 |
-
} catch (e) {
|
| 154 |
-
setError("Could not access webcam: " + (e instanceof Error ? e.message : String(e)));
|
| 155 |
-
setWebcamActive(false);
|
| 156 |
-
}
|
| 157 |
-
};
|
| 158 |
-
setupWebcam();
|
| 159 |
-
return () => {
|
| 160 |
-
if (webcamStreamRef.current) {
|
| 161 |
-
webcamStreamRef.current.getTracks().forEach((track: MediaStreamTrack) => track.stop());
|
| 162 |
-
webcamStreamRef.current = null;
|
| 163 |
-
}
|
| 164 |
-
setWebcamActive(false);
|
| 165 |
-
};
|
| 166 |
-
}, [mode]);
|
| 167 |
-
|
| 168 |
-
// Webcam mode: process frames with setInterval
|
| 169 |
-
useEffect(() => {
|
| 170 |
-
if (mode !== "Webcam" || !isLoaded || !webcamActive) return;
|
| 171 |
-
let interval: ReturnType<typeof setInterval> | null = null;
|
| 172 |
-
interval = setInterval(() => {
|
| 173 |
-
processVideoFrame();
|
| 174 |
-
}, 1000);
|
| 175 |
-
return () => {
|
| 176 |
-
if (interval) clearInterval(interval);
|
| 177 |
-
};
|
| 178 |
-
}, [mode, isLoaded, prompt, runInference, webcamActive]);
|
| 179 |
-
|
| 180 |
-
// URL mode: process frames with setInterval
|
| 181 |
-
useEffect(() => {
|
| 182 |
-
if (mode !== "URL" || !isLoaded || !urlProcessing) return;
|
| 183 |
-
let interval: ReturnType<typeof setInterval> | null = null;
|
| 184 |
-
interval = setInterval(() => {
|
| 185 |
-
processVideoFrame();
|
| 186 |
-
}, 1000);
|
| 187 |
-
return () => {
|
| 188 |
-
if (interval) clearInterval(interval);
|
| 189 |
-
};
|
| 190 |
-
}, [mode, isLoaded, prompt, runInference, urlProcessing]);
|
| 191 |
-
|
| 192 |
-
// File video mode: process frames with setInterval
|
| 193 |
-
useEffect(() => {
|
| 194 |
-
if (mode !== "File" || !isLoaded || !uploadedFile || !isVideoFile(uploadedFile) || !videoProcessing) return;
|
| 195 |
-
let interval: ReturnType<typeof setInterval> | null = null;
|
| 196 |
-
interval = setInterval(() => {
|
| 197 |
-
processVideoFrame();
|
| 198 |
-
}, 1000);
|
| 199 |
-
return () => {
|
| 200 |
-
if (interval) clearInterval(interval);
|
| 201 |
-
};
|
| 202 |
-
}, [mode, isLoaded, prompt, runInference, uploadedFile, videoProcessing]);
|
| 203 |
-
|
| 204 |
-
// Example video mode: process frames with setInterval
|
| 205 |
-
useEffect(() => {
|
| 206 |
-
if (mode !== "File" || uploadedFile || !isLoaded || !exampleProcessing) return;
|
| 207 |
-
let interval: ReturnType<typeof setInterval> | null = null;
|
| 208 |
-
interval = setInterval(() => {
|
| 209 |
-
processVideoFrame();
|
| 210 |
-
}, 1000);
|
| 211 |
-
return () => {
|
| 212 |
-
if (interval) clearInterval(interval);
|
| 213 |
-
};
|
| 214 |
-
}, [mode, isLoaded, prompt, runInference, uploadedFile, exampleProcessing]);
|
| 215 |
-
|
| 216 |
-
// File mode: process uploaded image (only on button click)
|
| 217 |
-
const handleProcessImage = async () => {
|
| 218 |
-
if (!isLoaded || !uploadedFile || !isImageFile(uploadedFile) || !imageRef.current || !canvasRef.current) return;
|
| 219 |
-
const img = imageRef.current;
|
| 220 |
-
const canvas = canvasRef.current;
|
| 221 |
-
canvas.width = img.naturalWidth;
|
| 222 |
-
canvas.height = img.naturalHeight;
|
| 223 |
-
setCanvasDims({w:canvas.width,h:canvas.height});
|
| 224 |
-
setVideoDims({w:img.naturalWidth,h:img.naturalHeight});
|
| 225 |
-
const ctx = canvas.getContext("2d");
|
| 226 |
-
if (!ctx) return;
|
| 227 |
-
ctx.drawImage(img, 0, 0, canvas.width, canvas.height);
|
| 228 |
-
setProcessing(true);
|
| 229 |
-
setError(null);
|
| 230 |
-
setInferenceStatus("Running inference...");
|
| 231 |
-
await runInference(img, prompt, (output: string) => {
|
| 232 |
-
setDebugOutput(output);
|
| 233 |
-
setInferenceStatus("Inference complete.");
|
| 234 |
-
ctx.drawImage(img, 0, 0, canvas.width, canvas.height);
|
| 235 |
-
let boxes = extractJsonFromMarkdown(output) || [];
|
| 236 |
-
if (boxes.length === 0 && Array.isArray(output)) {
|
| 237 |
-
boxes = parseFlatBoxArray(output);
|
| 238 |
-
}
|
| 239 |
-
boxes = normalizeBoxes(boxes);
|
| 240 |
-
console.log("Model output:", output);
|
| 241 |
-
console.log("Boxes after normalization:", boxes);
|
| 242 |
-
console.log("Canvas size:", canvas.width, canvas.height);
|
| 243 |
-
if (boxes.length > 0) {
|
| 244 |
-
const [x1, y1, x2, y2] = boxes[0].bbox_2d;
|
| 245 |
-
console.log("First box coords:", x1, y1, x2, y2);
|
| 246 |
-
}
|
| 247 |
-
if (boxes.length === 0) setInferenceStatus("No boxes detected or model output invalid.");
|
| 248 |
-
if (Array.isArray(boxes) && boxes.length > 0) {
|
| 249 |
-
const scaleX = canvas.width / img.naturalWidth;
|
| 250 |
-
const scaleY = canvas.height / img.naturalHeight;
|
| 251 |
-
drawBoundingBoxesOnCanvas(ctx, boxes, { scaleX, scaleY });
|
| 252 |
-
}
|
| 253 |
-
setImageProcessed(true);
|
| 254 |
-
});
|
| 255 |
-
setProcessing(false);
|
| 256 |
-
};
|
| 257 |
-
|
| 258 |
-
// File mode: process uploaded video frames (start/stop)
|
| 259 |
-
const handleToggleVideoProcessing = () => {
|
| 260 |
-
setVideoProcessing((prev) => !prev);
|
| 261 |
-
};
|
| 262 |
-
|
| 263 |
-
// Handle start/stop for example video processing
|
| 264 |
-
const handleToggleExampleProcessing = () => {
|
| 265 |
-
setExampleProcessing((prev) => !prev);
|
| 266 |
-
};
|
| 267 |
-
|
| 268 |
-
// Handle start/stop for URL video processing
|
| 269 |
-
const handleToggleUrlProcessing = () => {
|
| 270 |
-
setUrlProcessing((prev) => !prev);
|
| 271 |
-
};
|
| 272 |
-
|
| 273 |
-
// Test draw box function
|
| 274 |
-
const handleTestDrawBox = () => {
|
| 275 |
-
if (!canvasRef.current) return;
|
| 276 |
-
const canvas = canvasRef.current;
|
| 277 |
-
const ctx = canvas.getContext("2d");
|
| 278 |
-
if (!ctx) return;
|
| 279 |
-
ctx.clearRect(0, 0, canvas.width, canvas.height);
|
| 280 |
-
ctx.strokeStyle = "#FF00FF";
|
| 281 |
-
ctx.lineWidth = 4;
|
| 282 |
-
ctx.strokeRect(40, 40, Math.max(40,canvas.width/4), Math.max(40,canvas.height/4));
|
| 283 |
-
ctx.font = "20px Arial";
|
| 284 |
-
ctx.fillStyle = "#FF00FF";
|
| 285 |
-
ctx.fillText("Test Box", 50, 35);
|
| 286 |
-
};
|
| 287 |
-
|
| 288 |
-
return (
|
| 289 |
-
<div className="absolute inset-0 text-white">
|
| 290 |
-
<div className="fixed top-0 left-0 w-full bg-gray-900 text-white text-center py-2 z-50">
|
| 291 |
-
{isLoading ? "Loading model..." : isLoaded ? "Model loaded" : modelError ? `Model error: ${modelError}` : "Model not loaded"}
|
| 292 |
-
</div>
|
| 293 |
-
<div className="text-center text-sm text-blue-300 mt-2">{inferenceStatus}</div>
|
| 294 |
-
<div className="flex flex-col items-center justify-center h-full w-full">
|
| 295 |
-
{/* Mode Selector */}
|
| 296 |
-
<div className="mb-6">
|
| 297 |
-
<div className="flex space-x-4">
|
| 298 |
-
{MODES.map((m) => (
|
| 299 |
-
<button
|
| 300 |
-
key={m}
|
| 301 |
-
className={`px-6 py-2 rounded-lg font-semibold transition-all duration-200 ${
|
| 302 |
-
mode === m ? "bg-blue-600 text-white" : "bg-gray-700 text-gray-300 hover:bg-blue-500"
|
| 303 |
-
}`}
|
| 304 |
-
onClick={() => setMode(m)}
|
| 305 |
-
>
|
| 306 |
-
{m}
|
| 307 |
-
</button>
|
| 308 |
-
))}
|
| 309 |
-
</div>
|
| 310 |
-
</div>
|
| 311 |
-
|
| 312 |
-
{/* Mode Content */}
|
| 313 |
-
<div className="w-full max-w-2xl flex-1 flex flex-col items-center justify-center">
|
| 314 |
-
{mode === "Webcam" && (
|
| 315 |
-
<div className="w-full text-center flex flex-col items-center">
|
| 316 |
-
<div className="mb-4 w-full max-w-xl">
|
| 317 |
-
<label className="block text-left mb-2 font-medium">Detection Prompt:</label>
|
| 318 |
-
<textarea
|
| 319 |
-
className="w-full p-2 rounded-lg text-black"
|
| 320 |
-
rows={3}
|
| 321 |
-
value={prompt}
|
| 322 |
-
onChange={(e) => setPrompt(e.target.value)}
|
| 323 |
-
/>
|
| 324 |
-
</div>
|
| 325 |
-
<div className="relative w-full max-w-xl">
|
| 326 |
-
<video
|
| 327 |
-
ref={videoRef}
|
| 328 |
-
autoPlay
|
| 329 |
-
muted
|
| 330 |
-
playsInline
|
| 331 |
-
className="w-full rounded-lg shadow-lg mb-2"
|
| 332 |
-
style={{ background: "#222" }}
|
| 333 |
-
/>
|
| 334 |
-
<canvas
|
| 335 |
-
ref={canvasRef}
|
| 336 |
-
className="absolute top-0 left-0 w-full h-full pointer-events-none"
|
| 337 |
-
style={{ zIndex: 10, pointerEvents: "none" }}
|
| 338 |
-
/>
|
| 339 |
-
</div>
|
| 340 |
-
{processing && <div className="text-blue-400 mt-2">Processing frame...</div>}
|
| 341 |
-
{error && <div className="text-red-400 mt-2">Error: {error}</div>}
|
| 342 |
-
</div>
|
| 343 |
-
)}
|
| 344 |
-
{mode === "URL" && (
|
| 345 |
-
<div className="w-full text-center flex flex-col items-center">
|
| 346 |
-
<p className="mb-4">Enter a video stream URL (e.g., HTTP MP4, MJPEG, HLS, etc.):</p>
|
| 347 |
-
<div className="flex w-full max-w-xl mb-4">
|
| 348 |
-
<input
|
| 349 |
-
type="text"
|
| 350 |
-
className="flex-1 px-4 py-2 rounded-l-lg text-black"
|
| 351 |
-
value={inputUrl}
|
| 352 |
-
onChange={(e) => setInputUrl(e.target.value)}
|
| 353 |
-
placeholder="Paste video URL here"
|
| 354 |
-
/>
|
| 355 |
-
<button
|
| 356 |
-
className="px-4 py-2 rounded-r-lg bg-blue-600 text-white font-semibold"
|
| 357 |
-
onClick={() => setVideoUrl(inputUrl)}
|
| 358 |
-
>
|
| 359 |
-
Load
|
| 360 |
-
</button>
|
| 361 |
-
</div>
|
| 362 |
-
<div className="mb-4 w-full max-w-xl">
|
| 363 |
-
<label className="block text-left mb-2 font-medium">Detection Prompt:</label>
|
| 364 |
-
<textarea
|
| 365 |
-
className="w-full p-2 rounded-lg text-black"
|
| 366 |
-
rows={3}
|
| 367 |
-
value={prompt}
|
| 368 |
-
onChange={(e) => setPrompt(e.target.value)}
|
| 369 |
-
/>
|
| 370 |
-
</div>
|
| 371 |
-
<div className="relative w-full max-w-xl">
|
| 372 |
-
<video
|
| 373 |
-
ref={videoRef}
|
| 374 |
-
src={videoUrl}
|
| 375 |
-
controls
|
| 376 |
-
autoPlay
|
| 377 |
-
loop
|
| 378 |
-
className="w-full rounded-lg shadow-lg mb-2"
|
| 379 |
-
style={{ background: "#222" }}
|
| 380 |
-
/>
|
| 381 |
-
<canvas
|
| 382 |
-
ref={canvasRef}
|
| 383 |
-
className="absolute top-0 left-0 w-full h-full pointer-events-none"
|
| 384 |
-
style={{ zIndex: 10, pointerEvents: "none" }}
|
| 385 |
-
/>
|
| 386 |
-
<button
|
| 387 |
-
className="mt-4 px-6 py-2 rounded-lg bg-blue-600 text-white font-semibold"
|
| 388 |
-
onClick={handleToggleUrlProcessing}
|
| 389 |
-
>
|
| 390 |
-
{urlProcessing ? "Stop Processing" : "Start Processing"}
|
| 391 |
-
</button>
|
| 392 |
-
</div>
|
| 393 |
-
{processing && <div className="text-blue-400 mt-2">Processing frame...</div>}
|
| 394 |
-
{error && <div className="text-red-400 mt-2">Error: {error}</div>}
|
| 395 |
-
<button
|
| 396 |
-
className="mt-4 px-6 py-2 rounded-lg bg-gray-600 text-white font-semibold"
|
| 397 |
-
onClick={handleTestDrawBox}
|
| 398 |
-
>
|
| 399 |
-
Test Draw Box
|
| 400 |
-
</button>
|
| 401 |
-
<div className="mt-2 p-2 bg-gray-800 rounded text-xs">
|
| 402 |
-
<div>Canvas: {canvasDims ? `${canvasDims.w}x${canvasDims.h}` : "-"} | Video: {videoDims ? `${videoDims.w}x${videoDims.h}` : "-"}</div>
|
| 403 |
-
<div>Raw Model Output:</div>
|
| 404 |
-
<pre className="overflow-x-auto max-h-32 whitespace-pre-wrap">{debugOutput}</pre>
|
| 405 |
-
</div>
|
| 406 |
-
</div>
|
| 407 |
-
)}
|
| 408 |
-
{mode === "File" && (
|
| 409 |
-
<div className="w-full text-center flex flex-col items-center">
|
| 410 |
-
<div className="mb-4 w-full max-w-xl">
|
| 411 |
-
<label className="block text-left mb-2 font-medium">Detection Prompt:</label>
|
| 412 |
-
<textarea
|
| 413 |
-
className="w-full p-2 rounded-lg text-black"
|
| 414 |
-
rows={3}
|
| 415 |
-
value={prompt}
|
| 416 |
-
onChange={(e) => setPrompt(e.target.value)}
|
| 417 |
-
/>
|
| 418 |
-
</div>
|
| 419 |
-
<div className="mb-4 w-full max-w-xl">
|
| 420 |
-
<input
|
| 421 |
-
type="file"
|
| 422 |
-
accept="image/*,video/*"
|
| 423 |
-
onChange={handleFileChange}
|
| 424 |
-
className="block w-full text-sm text-gray-300 file:mr-4 file:py-2 file:px-4 file:rounded-lg file:border-0 file:text-sm file:font-semibold file:bg-blue-600 file:text-white hover:file:bg-blue-700"
|
| 425 |
-
/>
|
| 426 |
-
</div>
|
| 427 |
-
{/* Show uploaded image */}
|
| 428 |
-
{uploadedFile && isImageFile(uploadedFile) && (
|
| 429 |
-
<div className="relative w-full max-w-xl">
|
| 430 |
-
<img
|
| 431 |
-
ref={imageRef}
|
| 432 |
-
src={uploadedUrl}
|
| 433 |
-
alt="Uploaded"
|
| 434 |
-
className="w-full rounded-lg shadow-lg mb-2"
|
| 435 |
-
style={{ background: "#222" }}
|
| 436 |
-
/>
|
| 437 |
-
<canvas
|
| 438 |
-
ref={canvasRef}
|
| 439 |
-
className="absolute top-0 left-0 w-full h-full pointer-events-none"
|
| 440 |
-
style={{ zIndex: 10, pointerEvents: "none" }}
|
| 441 |
-
/>
|
| 442 |
-
<button
|
| 443 |
-
className="mt-4 px-6 py-2 rounded-lg bg-blue-600 text-white font-semibold"
|
| 444 |
-
onClick={handleProcessImage}
|
| 445 |
-
disabled={processing}
|
| 446 |
-
>
|
| 447 |
-
{processing ? "Processing..." : imageProcessed ? "Reprocess Image" : "Process Image"}
|
| 448 |
-
</button>
|
| 449 |
-
</div>
|
| 450 |
-
)}
|
| 451 |
-
{/* Show uploaded video */}
|
| 452 |
-
{uploadedFile && isVideoFile(uploadedFile) && (
|
| 453 |
-
<div className="relative w-full max-w-xl">
|
| 454 |
-
<video
|
| 455 |
-
ref={videoRef}
|
| 456 |
-
src={uploadedUrl}
|
| 457 |
-
controls
|
| 458 |
-
autoPlay
|
| 459 |
-
loop
|
| 460 |
-
className="w-full rounded-lg shadow-lg mb-2"
|
| 461 |
-
style={{ background: "#222" }}
|
| 462 |
-
/>
|
| 463 |
-
<canvas
|
| 464 |
-
ref={canvasRef}
|
| 465 |
-
className="absolute top-0 left-0 w-full h-full pointer-events-none"
|
| 466 |
-
style={{ zIndex: 10, pointerEvents: "none" }}
|
| 467 |
-
/>
|
| 468 |
-
<button
|
| 469 |
-
className="mt-4 px-6 py-2 rounded-lg bg-blue-600 text-white font-semibold"
|
| 470 |
-
onClick={handleToggleVideoProcessing}
|
| 471 |
-
>
|
| 472 |
-
{videoProcessing ? "Stop Processing" : "Start Processing"}
|
| 473 |
-
</button>
|
| 474 |
-
</div>
|
| 475 |
-
)}
|
| 476 |
-
{/* Show example video if no file uploaded */}
|
| 477 |
-
{!uploadedFile && (
|
| 478 |
-
<div className="relative w-full max-w-xl">
|
| 479 |
-
<video
|
| 480 |
-
ref={videoRef}
|
| 481 |
-
src={EXAMPLE_VIDEO_URL}
|
| 482 |
-
controls
|
| 483 |
-
autoPlay
|
| 484 |
-
loop
|
| 485 |
-
className="w-full rounded-lg shadow-lg mb-2"
|
| 486 |
-
style={{ background: "#222" }}
|
| 487 |
-
/>
|
| 488 |
-
<canvas
|
| 489 |
-
ref={canvasRef}
|
| 490 |
-
className="absolute top-0 left-0 w-full h-full pointer-events-none"
|
| 491 |
-
style={{ zIndex: 10, pointerEvents: "none" }}
|
| 492 |
-
/>
|
| 493 |
-
<button
|
| 494 |
-
className="mt-4 px-6 py-2 rounded-lg bg-blue-600 text-white font-semibold"
|
| 495 |
-
onClick={handleToggleExampleProcessing}
|
| 496 |
-
>
|
| 497 |
-
{exampleProcessing ? "Stop Processing" : "Start Processing"}
|
| 498 |
-
</button>
|
| 499 |
-
</div>
|
| 500 |
-
)}
|
| 501 |
-
{processing && <div className="text-blue-400 mt-2">Processing frame...</div>}
|
| 502 |
-
{error && <div className="text-red-400 mt-2">Error: {error}</div>}
|
| 503 |
-
<button
|
| 504 |
-
className="mt-4 px-6 py-2 rounded-lg bg-gray-600 text-white font-semibold"
|
| 505 |
-
onClick={handleTestDrawBox}
|
| 506 |
-
>
|
| 507 |
-
Test Draw Box
|
| 508 |
-
</button>
|
| 509 |
-
<div className="mt-2 p-2 bg-gray-800 rounded text-xs">
|
| 510 |
-
<div>Canvas: {canvasDims ? `${canvasDims.w}x${canvasDims.h}` : "-"} | Video: {videoDims ? `${videoDims.w}x${videoDims.h}` : "-"}</div>
|
| 511 |
-
<div>Raw Model Output:</div>
|
| 512 |
-
<pre className="overflow-x-auto max-h-32 whitespace-pre-wrap">{debugOutput}</pre>
|
| 513 |
-
</div>
|
| 514 |
-
</div>
|
| 515 |
-
)}
|
| 516 |
-
</div>
|
| 517 |
-
</div>
|
| 518 |
-
</div>
|
| 519 |
-
);
|
| 520 |
}
|
|
|
|
| 1 |
+
import { useState, useRef, useEffect } from "react";
|
| 2 |
+
import { useVLMContext } from "../context/useVLMContext";
|
| 3 |
+
import { extractJsonFromMarkdown, drawBoundingBoxesOnCanvas } from "./BoxAnnotator";
|
| 4 |
+
|
| 5 |
+
const MODES = ["Webcam", "URL", "File"] as const;
|
| 6 |
+
type Mode = typeof MODES[number];
|
| 7 |
+
|
| 8 |
+
const EXAMPLE_VIDEO_URL = "https://huggingface.co/spaces/Quazim0t0/test/resolve/main/videos/1.mp4";
|
| 9 |
+
const EXAMPLE_PROMPT = "Detect all people in the image. For each person, output a JSON array of objects with fields: 'label' (string) and 'bbox_2d' ([x1, y1, x2, y2]) where coordinates are in pixel values. Example: [{\"label\": \"person\", \"bbox_2d\": [100, 50, 200, 300]}]";
|
| 10 |
+
|
| 11 |
+
function parseFlatBoxArray(arr: any[]): { label: string, bbox_2d: number[] }[] {
|
| 12 |
+
if (typeof arr[0] === "string" && Array.isArray(arr[1])) {
|
| 13 |
+
const label = arr[0];
|
| 14 |
+
return arr.slice(1).map(bbox => ({ label, bbox_2d: bbox }));
|
| 15 |
+
}
|
| 16 |
+
return [];
|
| 17 |
+
}
|
| 18 |
+
|
| 19 |
+
function normalizeBoxes(raw: any): { label: string, bbox_2d: number[] }[] {
|
| 20 |
+
if (!raw) return [];
|
| 21 |
+
let boxes = [];
|
| 22 |
+
if (typeof raw === "object" && raw !== null && Array.isArray(raw.image)) {
|
| 23 |
+
boxes = raw.image;
|
| 24 |
+
} else if (Array.isArray(raw)) {
|
| 25 |
+
boxes = raw;
|
| 26 |
+
} else if (typeof raw === "object" && raw !== null) {
|
| 27 |
+
boxes = [raw];
|
| 28 |
+
}
|
| 29 |
+
return boxes
|
| 30 |
+
.map((obj: any) => {
|
| 31 |
+
if (!obj || !obj.bbox_2d) return null;
|
| 32 |
+
let bbox = obj.bbox_2d;
|
| 33 |
+
// If bbox_2d is [[x1, y1], [x2, y2]], convert to [x1, y1, x2, y2]
|
| 34 |
+
if (
|
| 35 |
+
Array.isArray(bbox) &&
|
| 36 |
+
bbox.length === 2 &&
|
| 37 |
+
Array.isArray(bbox[0]) &&
|
| 38 |
+
Array.isArray(bbox[1]) &&
|
| 39 |
+
bbox[0].length === 2 &&
|
| 40 |
+
bbox[1].length === 2
|
| 41 |
+
) {
|
| 42 |
+
bbox = [bbox[0][0], bbox[0][1], bbox[1][0], bbox[1][1]];
|
| 43 |
+
}
|
| 44 |
+
// If bbox_2d is [x1, y1, x2, y2], use as-is
|
| 45 |
+
if (
|
| 46 |
+
Array.isArray(bbox) &&
|
| 47 |
+
bbox.length === 4 &&
|
| 48 |
+
bbox.every((v: any) => typeof v === "number")
|
| 49 |
+
) {
|
| 50 |
+
return { ...obj, bbox_2d: bbox };
|
| 51 |
+
}
|
| 52 |
+
// Otherwise, skip
|
| 53 |
+
return null;
|
| 54 |
+
})
|
| 55 |
+
.filter((obj: any) => obj);
|
| 56 |
+
}
|
| 57 |
+
|
| 58 |
+
function isImageFile(file: File) {
|
| 59 |
+
return file.type.startsWith("image/");
|
| 60 |
+
}
|
| 61 |
+
function isVideoFile(file: File) {
|
| 62 |
+
return file.type.startsWith("video/");
|
| 63 |
+
}
|
| 64 |
+
|
| 65 |
+
export default function MultiSourceCaptioningView() {
|
| 66 |
+
const [mode, setMode] = useState<Mode>("File");
|
| 67 |
+
const [videoUrl, setVideoUrl] = useState<string>(EXAMPLE_VIDEO_URL);
|
| 68 |
+
const [inputUrl, setInputUrl] = useState<string>(EXAMPLE_VIDEO_URL);
|
| 69 |
+
const [prompt, setPrompt] = useState<string>(EXAMPLE_PROMPT);
|
| 70 |
+
const [processing, setProcessing] = useState(false);
|
| 71 |
+
const [error, setError] = useState<string | null>(null);
|
| 72 |
+
const [webcamActive, setWebcamActive] = useState(false);
|
| 73 |
+
const [uploadedFile, setUploadedFile] = useState<File | null>(null);
|
| 74 |
+
const [uploadedUrl, setUploadedUrl] = useState<string>("");
|
| 75 |
+
const [videoProcessing, setVideoProcessing] = useState(false);
|
| 76 |
+
const [imageProcessed, setImageProcessed] = useState(false);
|
| 77 |
+
const [exampleProcessing, setExampleProcessing] = useState(false);
|
| 78 |
+
const [urlProcessing, setUrlProcessing] = useState(false);
|
| 79 |
+
const [debugOutput, setDebugOutput] = useState<string>("");
|
| 80 |
+
const [canvasDims, setCanvasDims] = useState<{w:number,h:number}|null>(null);
|
| 81 |
+
const [videoDims, setVideoDims] = useState<{w:number,h:number}|null>(null);
|
| 82 |
+
const [inferenceStatus, setInferenceStatus] = useState<string>("");
|
| 83 |
+
|
| 84 |
+
const videoRef = useRef<HTMLVideoElement | null>(null);
|
| 85 |
+
const canvasRef = useRef<HTMLCanvasElement | null>(null);
|
| 86 |
+
const imageRef = useRef<HTMLImageElement | null>(null);
|
| 87 |
+
const webcamStreamRef = useRef<MediaStream | null>(null);
|
| 88 |
+
const { isLoaded, isLoading, error: modelError, runInference } = useVLMContext();
|
| 89 |
+
|
| 90 |
+
const processVideoFrame = async () => {
|
| 91 |
+
if (!videoRef.current || !canvasRef.current) return;
|
| 92 |
+
const video = videoRef.current;
|
| 93 |
+
const canvas = canvasRef.current;
|
| 94 |
+
if (video.paused || video.ended || video.videoWidth === 0) return;
|
| 95 |
+
canvas.width = video.videoWidth;
|
| 96 |
+
canvas.height = video.videoHeight;
|
| 97 |
+
const ctx = canvas.getContext("2d");
|
| 98 |
+
if (!ctx) return;
|
| 99 |
+
ctx.drawImage(video, 0, 0, canvas.width, canvas.height);
|
| 100 |
+
await runInference(video, prompt, (output: string) => {
|
| 101 |
+
setDebugOutput(output); // <-- Ensure Raw Model Output is updated
|
| 102 |
+
let boxes = extractJsonFromMarkdown(output) || [];
|
| 103 |
+
if (boxes.length === 0 && Array.isArray(output)) {
|
| 104 |
+
boxes = parseFlatBoxArray(output);
|
| 105 |
+
}
|
| 106 |
+
boxes = normalizeBoxes(boxes);
|
| 107 |
+
console.log("Model output:", output);
|
| 108 |
+
console.log("Boxes after normalization:", boxes);
|
| 109 |
+
console.log("Canvas size:", canvas.width, canvas.height);
|
| 110 |
+
if (boxes.length > 0) {
|
| 111 |
+
const [x1, y1, x2, y2] = boxes[0].bbox_2d;
|
| 112 |
+
console.log("First box coords:", x1, y1, x2, y2);
|
| 113 |
+
}
|
| 114 |
+
if (boxes.length === 0) setInferenceStatus("No boxes detected or model output invalid.");
|
| 115 |
+
if (Array.isArray(boxes) && boxes.length > 0) {
|
| 116 |
+
const scaleX = canvas.width / video.videoWidth;
|
| 117 |
+
const scaleY = canvas.height / video.videoHeight;
|
| 118 |
+
ctx.clearRect(0, 0, canvas.width, canvas.height); // Clear canvas before drawing boxes
|
| 119 |
+
drawBoundingBoxesOnCanvas(ctx, boxes, { color: "#FF00FF", lineWidth: 4, font: "20px Arial", scaleX, scaleY }); // Use visible color and thick line
|
| 120 |
+
}
|
| 121 |
+
});
|
| 122 |
+
};
|
| 123 |
+
|
| 124 |
+
const handleFileChange = (e: React.ChangeEvent<HTMLInputElement>) => {
|
| 125 |
+
const file = e.target.files?.[0] || null;
|
| 126 |
+
setUploadedFile(file);
|
| 127 |
+
setUploadedUrl(file ? URL.createObjectURL(file) : "");
|
| 128 |
+
setError(null);
|
| 129 |
+
setImageProcessed(false);
|
| 130 |
+
setVideoProcessing(false);
|
| 131 |
+
setExampleProcessing(false);
|
| 132 |
+
};
|
| 133 |
+
|
| 134 |
+
// Webcam setup and teardown (unchanged)
|
| 135 |
+
useEffect(() => {
|
| 136 |
+
if (mode !== "Webcam") {
|
| 137 |
+
if (webcamStreamRef.current) {
|
| 138 |
+
webcamStreamRef.current.getTracks().forEach((track: MediaStreamTrack) => track.stop());
|
| 139 |
+
webcamStreamRef.current = null;
|
| 140 |
+
}
|
| 141 |
+
setWebcamActive(false);
|
| 142 |
+
return;
|
| 143 |
+
}
|
| 144 |
+
const setupWebcam = async () => {
|
| 145 |
+
try {
|
| 146 |
+
setError(null);
|
| 147 |
+
const stream = await navigator.mediaDevices.getUserMedia({ video: true });
|
| 148 |
+
webcamStreamRef.current = stream;
|
| 149 |
+
if (videoRef.current) {
|
| 150 |
+
videoRef.current.srcObject = stream;
|
| 151 |
+
setWebcamActive(true);
|
| 152 |
+
}
|
| 153 |
+
} catch (e) {
|
| 154 |
+
setError("Could not access webcam: " + (e instanceof Error ? e.message : String(e)));
|
| 155 |
+
setWebcamActive(false);
|
| 156 |
+
}
|
| 157 |
+
};
|
| 158 |
+
setupWebcam();
|
| 159 |
+
return () => {
|
| 160 |
+
if (webcamStreamRef.current) {
|
| 161 |
+
webcamStreamRef.current.getTracks().forEach((track: MediaStreamTrack) => track.stop());
|
| 162 |
+
webcamStreamRef.current = null;
|
| 163 |
+
}
|
| 164 |
+
setWebcamActive(false);
|
| 165 |
+
};
|
| 166 |
+
}, [mode]);
|
| 167 |
+
|
| 168 |
+
// Webcam mode: process frames with setInterval
|
| 169 |
+
useEffect(() => {
|
| 170 |
+
if (mode !== "Webcam" || !isLoaded || !webcamActive) return;
|
| 171 |
+
let interval: ReturnType<typeof setInterval> | null = null;
|
| 172 |
+
interval = setInterval(() => {
|
| 173 |
+
processVideoFrame();
|
| 174 |
+
}, 1000);
|
| 175 |
+
return () => {
|
| 176 |
+
if (interval) clearInterval(interval);
|
| 177 |
+
};
|
| 178 |
+
}, [mode, isLoaded, prompt, runInference, webcamActive]);
|
| 179 |
+
|
| 180 |
+
// URL mode: process frames with setInterval
|
| 181 |
+
useEffect(() => {
|
| 182 |
+
if (mode !== "URL" || !isLoaded || !urlProcessing) return;
|
| 183 |
+
let interval: ReturnType<typeof setInterval> | null = null;
|
| 184 |
+
interval = setInterval(() => {
|
| 185 |
+
processVideoFrame();
|
| 186 |
+
}, 1000);
|
| 187 |
+
return () => {
|
| 188 |
+
if (interval) clearInterval(interval);
|
| 189 |
+
};
|
| 190 |
+
}, [mode, isLoaded, prompt, runInference, urlProcessing]);
|
| 191 |
+
|
| 192 |
+
// File video mode: process frames with setInterval
|
| 193 |
+
useEffect(() => {
|
| 194 |
+
if (mode !== "File" || !isLoaded || !uploadedFile || !isVideoFile(uploadedFile) || !videoProcessing) return;
|
| 195 |
+
let interval: ReturnType<typeof setInterval> | null = null;
|
| 196 |
+
interval = setInterval(() => {
|
| 197 |
+
processVideoFrame();
|
| 198 |
+
}, 1000);
|
| 199 |
+
return () => {
|
| 200 |
+
if (interval) clearInterval(interval);
|
| 201 |
+
};
|
| 202 |
+
}, [mode, isLoaded, prompt, runInference, uploadedFile, videoProcessing]);
|
| 203 |
+
|
| 204 |
+
// Example video mode: process frames with setInterval
|
| 205 |
+
useEffect(() => {
|
| 206 |
+
if (mode !== "File" || uploadedFile || !isLoaded || !exampleProcessing) return;
|
| 207 |
+
let interval: ReturnType<typeof setInterval> | null = null;
|
| 208 |
+
interval = setInterval(() => {
|
| 209 |
+
processVideoFrame();
|
| 210 |
+
}, 1000);
|
| 211 |
+
return () => {
|
| 212 |
+
if (interval) clearInterval(interval);
|
| 213 |
+
};
|
| 214 |
+
}, [mode, isLoaded, prompt, runInference, uploadedFile, exampleProcessing]);
|
| 215 |
+
|
| 216 |
+
// File mode: process uploaded image (only on button click)
|
| 217 |
+
const handleProcessImage = async () => {
|
| 218 |
+
if (!isLoaded || !uploadedFile || !isImageFile(uploadedFile) || !imageRef.current || !canvasRef.current) return;
|
| 219 |
+
const img = imageRef.current;
|
| 220 |
+
const canvas = canvasRef.current;
|
| 221 |
+
canvas.width = img.naturalWidth;
|
| 222 |
+
canvas.height = img.naturalHeight;
|
| 223 |
+
setCanvasDims({w:canvas.width,h:canvas.height});
|
| 224 |
+
setVideoDims({w:img.naturalWidth,h:img.naturalHeight});
|
| 225 |
+
const ctx = canvas.getContext("2d");
|
| 226 |
+
if (!ctx) return;
|
| 227 |
+
ctx.drawImage(img, 0, 0, canvas.width, canvas.height);
|
| 228 |
+
setProcessing(true);
|
| 229 |
+
setError(null);
|
| 230 |
+
setInferenceStatus("Running inference...");
|
| 231 |
+
await runInference(img, prompt, (output: string) => {
|
| 232 |
+
setDebugOutput(output);
|
| 233 |
+
setInferenceStatus("Inference complete.");
|
| 234 |
+
ctx.drawImage(img, 0, 0, canvas.width, canvas.height);
|
| 235 |
+
let boxes = extractJsonFromMarkdown(output) || [];
|
| 236 |
+
if (boxes.length === 0 && Array.isArray(output)) {
|
| 237 |
+
boxes = parseFlatBoxArray(output);
|
| 238 |
+
}
|
| 239 |
+
boxes = normalizeBoxes(boxes);
|
| 240 |
+
console.log("Model output:", output);
|
| 241 |
+
console.log("Boxes after normalization:", boxes);
|
| 242 |
+
console.log("Canvas size:", canvas.width, canvas.height);
|
| 243 |
+
if (boxes.length > 0) {
|
| 244 |
+
const [x1, y1, x2, y2] = boxes[0].bbox_2d;
|
| 245 |
+
console.log("First box coords:", x1, y1, x2, y2);
|
| 246 |
+
}
|
| 247 |
+
if (boxes.length === 0) setInferenceStatus("No boxes detected or model output invalid.");
|
| 248 |
+
if (Array.isArray(boxes) && boxes.length > 0) {
|
| 249 |
+
const scaleX = canvas.width / img.naturalWidth;
|
| 250 |
+
const scaleY = canvas.height / img.naturalHeight;
|
| 251 |
+
drawBoundingBoxesOnCanvas(ctx, boxes, { scaleX, scaleY });
|
| 252 |
+
}
|
| 253 |
+
setImageProcessed(true);
|
| 254 |
+
});
|
| 255 |
+
setProcessing(false);
|
| 256 |
+
};
|
| 257 |
+
|
| 258 |
+
// File mode: process uploaded video frames (start/stop)
|
| 259 |
+
const handleToggleVideoProcessing = () => {
|
| 260 |
+
setVideoProcessing((prev) => !prev);
|
| 261 |
+
};
|
| 262 |
+
|
| 263 |
+
// Handle start/stop for example video processing
|
| 264 |
+
const handleToggleExampleProcessing = () => {
|
| 265 |
+
setExampleProcessing((prev) => !prev);
|
| 266 |
+
};
|
| 267 |
+
|
| 268 |
+
// Handle start/stop for URL video processing
|
| 269 |
+
const handleToggleUrlProcessing = () => {
|
| 270 |
+
setUrlProcessing((prev) => !prev);
|
| 271 |
+
};
|
| 272 |
+
|
| 273 |
+
// Test draw box function
|
| 274 |
+
const handleTestDrawBox = () => {
|
| 275 |
+
if (!canvasRef.current) return;
|
| 276 |
+
const canvas = canvasRef.current;
|
| 277 |
+
const ctx = canvas.getContext("2d");
|
| 278 |
+
if (!ctx) return;
|
| 279 |
+
ctx.clearRect(0, 0, canvas.width, canvas.height);
|
| 280 |
+
ctx.strokeStyle = "#FF00FF";
|
| 281 |
+
ctx.lineWidth = 4;
|
| 282 |
+
ctx.strokeRect(40, 40, Math.max(40,canvas.width/4), Math.max(40,canvas.height/4));
|
| 283 |
+
ctx.font = "20px Arial";
|
| 284 |
+
ctx.fillStyle = "#FF00FF";
|
| 285 |
+
ctx.fillText("Test Box", 50, 35);
|
| 286 |
+
};
|
| 287 |
+
|
| 288 |
+
return (
|
| 289 |
+
<div className="absolute inset-0 text-white">
|
| 290 |
+
<div className="fixed top-0 left-0 w-full bg-gray-900 text-white text-center py-2 z-50">
|
| 291 |
+
{isLoading ? "Loading model..." : isLoaded ? "Model loaded" : modelError ? `Model error: ${modelError}` : "Model not loaded"}
|
| 292 |
+
</div>
|
| 293 |
+
<div className="text-center text-sm text-blue-300 mt-2">{inferenceStatus}</div>
|
| 294 |
+
<div className="flex flex-col items-center justify-center h-full w-full">
|
| 295 |
+
{/* Mode Selector */}
|
| 296 |
+
<div className="mb-6">
|
| 297 |
+
<div className="flex space-x-4">
|
| 298 |
+
{MODES.map((m) => (
|
| 299 |
+
<button
|
| 300 |
+
key={m}
|
| 301 |
+
className={`px-6 py-2 rounded-lg font-semibold transition-all duration-200 ${
|
| 302 |
+
mode === m ? "bg-blue-600 text-white" : "bg-gray-700 text-gray-300 hover:bg-blue-500"
|
| 303 |
+
}`}
|
| 304 |
+
onClick={() => setMode(m)}
|
| 305 |
+
>
|
| 306 |
+
{m}
|
| 307 |
+
</button>
|
| 308 |
+
))}
|
| 309 |
+
</div>
|
| 310 |
+
</div>
|
| 311 |
+
|
| 312 |
+
{/* Mode Content */}
|
| 313 |
+
<div className="w-full max-w-2xl flex-1 flex flex-col items-center justify-center">
|
| 314 |
+
{mode === "Webcam" && (
|
| 315 |
+
<div className="w-full text-center flex flex-col items-center">
|
| 316 |
+
<div className="mb-4 w-full max-w-xl">
|
| 317 |
+
<label className="block text-left mb-2 font-medium">Detection Prompt:</label>
|
| 318 |
+
<textarea
|
| 319 |
+
className="w-full p-2 rounded-lg text-black"
|
| 320 |
+
rows={3}
|
| 321 |
+
value={prompt}
|
| 322 |
+
onChange={(e) => setPrompt(e.target.value)}
|
| 323 |
+
/>
|
| 324 |
+
</div>
|
| 325 |
+
<div className="relative w-full max-w-xl">
|
| 326 |
+
<video
|
| 327 |
+
ref={videoRef}
|
| 328 |
+
autoPlay
|
| 329 |
+
muted
|
| 330 |
+
playsInline
|
| 331 |
+
className="w-full rounded-lg shadow-lg mb-2"
|
| 332 |
+
style={{ background: "#222" }}
|
| 333 |
+
/>
|
| 334 |
+
<canvas
|
| 335 |
+
ref={canvasRef}
|
| 336 |
+
className="absolute top-0 left-0 w-full h-full pointer-events-none"
|
| 337 |
+
style={{ zIndex: 10, pointerEvents: "none" }}
|
| 338 |
+
/>
|
| 339 |
+
</div>
|
| 340 |
+
{processing && <div className="text-blue-400 mt-2">Processing frame...</div>}
|
| 341 |
+
{error && <div className="text-red-400 mt-2">Error: {error}</div>}
|
| 342 |
+
</div>
|
| 343 |
+
)}
|
| 344 |
+
{mode === "URL" && (
|
| 345 |
+
<div className="w-full text-center flex flex-col items-center">
|
| 346 |
+
<p className="mb-4">Enter a video stream URL (e.g., HTTP MP4, MJPEG, HLS, etc.):</p>
|
| 347 |
+
<div className="flex w-full max-w-xl mb-4">
|
| 348 |
+
<input
|
| 349 |
+
type="text"
|
| 350 |
+
className="flex-1 px-4 py-2 rounded-l-lg text-black"
|
| 351 |
+
value={inputUrl}
|
| 352 |
+
onChange={(e) => setInputUrl(e.target.value)}
|
| 353 |
+
placeholder="Paste video URL here"
|
| 354 |
+
/>
|
| 355 |
+
<button
|
| 356 |
+
className="px-4 py-2 rounded-r-lg bg-blue-600 text-white font-semibold"
|
| 357 |
+
onClick={() => setVideoUrl(inputUrl)}
|
| 358 |
+
>
|
| 359 |
+
Load
|
| 360 |
+
</button>
|
| 361 |
+
</div>
|
| 362 |
+
<div className="mb-4 w-full max-w-xl">
|
| 363 |
+
<label className="block text-left mb-2 font-medium">Detection Prompt:</label>
|
| 364 |
+
<textarea
|
| 365 |
+
className="w-full p-2 rounded-lg text-black"
|
| 366 |
+
rows={3}
|
| 367 |
+
value={prompt}
|
| 368 |
+
onChange={(e) => setPrompt(e.target.value)}
|
| 369 |
+
/>
|
| 370 |
+
</div>
|
| 371 |
+
<div className="relative w-full max-w-xl">
|
| 372 |
+
<video
|
| 373 |
+
ref={videoRef}
|
| 374 |
+
src={videoUrl}
|
| 375 |
+
controls
|
| 376 |
+
autoPlay
|
| 377 |
+
loop
|
| 378 |
+
className="w-full rounded-lg shadow-lg mb-2"
|
| 379 |
+
style={{ background: "#222" }}
|
| 380 |
+
/>
|
| 381 |
+
<canvas
|
| 382 |
+
ref={canvasRef}
|
| 383 |
+
className="absolute top-0 left-0 w-full h-full pointer-events-none"
|
| 384 |
+
style={{ zIndex: 10, pointerEvents: "none" }}
|
| 385 |
+
/>
|
| 386 |
+
<button
|
| 387 |
+
className="mt-4 px-6 py-2 rounded-lg bg-blue-600 text-white font-semibold"
|
| 388 |
+
onClick={handleToggleUrlProcessing}
|
| 389 |
+
>
|
| 390 |
+
{urlProcessing ? "Stop Processing" : "Start Processing"}
|
| 391 |
+
</button>
|
| 392 |
+
</div>
|
| 393 |
+
{processing && <div className="text-blue-400 mt-2">Processing frame...</div>}
|
| 394 |
+
{error && <div className="text-red-400 mt-2">Error: {error}</div>}
|
| 395 |
+
<button
|
| 396 |
+
className="mt-4 px-6 py-2 rounded-lg bg-gray-600 text-white font-semibold"
|
| 397 |
+
onClick={handleTestDrawBox}
|
| 398 |
+
>
|
| 399 |
+
Test Draw Box
|
| 400 |
+
</button>
|
| 401 |
+
<div className="mt-2 p-2 bg-gray-800 rounded text-xs">
|
| 402 |
+
<div>Canvas: {canvasDims ? `${canvasDims.w}x${canvasDims.h}` : "-"} | Video: {videoDims ? `${videoDims.w}x${videoDims.h}` : "-"}</div>
|
| 403 |
+
<div>Raw Model Output:</div>
|
| 404 |
+
<pre className="overflow-x-auto max-h-32 whitespace-pre-wrap">{debugOutput}</pre>
|
| 405 |
+
</div>
|
| 406 |
+
</div>
|
| 407 |
+
)}
|
| 408 |
+
{mode === "File" && (
|
| 409 |
+
<div className="w-full text-center flex flex-col items-center">
|
| 410 |
+
<div className="mb-4 w-full max-w-xl">
|
| 411 |
+
<label className="block text-left mb-2 font-medium">Detection Prompt:</label>
|
| 412 |
+
<textarea
|
| 413 |
+
className="w-full p-2 rounded-lg text-black"
|
| 414 |
+
rows={3}
|
| 415 |
+
value={prompt}
|
| 416 |
+
onChange={(e) => setPrompt(e.target.value)}
|
| 417 |
+
/>
|
| 418 |
+
</div>
|
| 419 |
+
<div className="mb-4 w-full max-w-xl">
|
| 420 |
+
<input
|
| 421 |
+
type="file"
|
| 422 |
+
accept="image/*,video/*"
|
| 423 |
+
onChange={handleFileChange}
|
| 424 |
+
className="block w-full text-sm text-gray-300 file:mr-4 file:py-2 file:px-4 file:rounded-lg file:border-0 file:text-sm file:font-semibold file:bg-blue-600 file:text-white hover:file:bg-blue-700"
|
| 425 |
+
/>
|
| 426 |
+
</div>
|
| 427 |
+
{/* Show uploaded image */}
|
| 428 |
+
{uploadedFile && isImageFile(uploadedFile) && (
|
| 429 |
+
<div className="relative w-full max-w-xl">
|
| 430 |
+
<img
|
| 431 |
+
ref={imageRef}
|
| 432 |
+
src={uploadedUrl}
|
| 433 |
+
alt="Uploaded"
|
| 434 |
+
className="w-full rounded-lg shadow-lg mb-2"
|
| 435 |
+
style={{ background: "#222" }}
|
| 436 |
+
/>
|
| 437 |
+
<canvas
|
| 438 |
+
ref={canvasRef}
|
| 439 |
+
className="absolute top-0 left-0 w-full h-full pointer-events-none"
|
| 440 |
+
style={{ zIndex: 10, pointerEvents: "none" }}
|
| 441 |
+
/>
|
| 442 |
+
<button
|
| 443 |
+
className="mt-4 px-6 py-2 rounded-lg bg-blue-600 text-white font-semibold"
|
| 444 |
+
onClick={handleProcessImage}
|
| 445 |
+
disabled={processing}
|
| 446 |
+
>
|
| 447 |
+
{processing ? "Processing..." : imageProcessed ? "Reprocess Image" : "Process Image"}
|
| 448 |
+
</button>
|
| 449 |
+
</div>
|
| 450 |
+
)}
|
| 451 |
+
{/* Show uploaded video */}
|
| 452 |
+
{uploadedFile && isVideoFile(uploadedFile) && (
|
| 453 |
+
<div className="relative w-full max-w-xl">
|
| 454 |
+
<video
|
| 455 |
+
ref={videoRef}
|
| 456 |
+
src={uploadedUrl}
|
| 457 |
+
controls
|
| 458 |
+
autoPlay
|
| 459 |
+
loop
|
| 460 |
+
className="w-full rounded-lg shadow-lg mb-2"
|
| 461 |
+
style={{ background: "#222" }}
|
| 462 |
+
/>
|
| 463 |
+
<canvas
|
| 464 |
+
ref={canvasRef}
|
| 465 |
+
className="absolute top-0 left-0 w-full h-full pointer-events-none"
|
| 466 |
+
style={{ zIndex: 10, pointerEvents: "none" }}
|
| 467 |
+
/>
|
| 468 |
+
<button
|
| 469 |
+
className="mt-4 px-6 py-2 rounded-lg bg-blue-600 text-white font-semibold"
|
| 470 |
+
onClick={handleToggleVideoProcessing}
|
| 471 |
+
>
|
| 472 |
+
{videoProcessing ? "Stop Processing" : "Start Processing"}
|
| 473 |
+
</button>
|
| 474 |
+
</div>
|
| 475 |
+
)}
|
| 476 |
+
{/* Show example video if no file uploaded */}
|
| 477 |
+
{!uploadedFile && (
|
| 478 |
+
<div className="relative w-full max-w-xl">
|
| 479 |
+
<video
|
| 480 |
+
ref={videoRef}
|
| 481 |
+
src={EXAMPLE_VIDEO_URL}
|
| 482 |
+
controls
|
| 483 |
+
autoPlay
|
| 484 |
+
loop
|
| 485 |
+
className="w-full rounded-lg shadow-lg mb-2"
|
| 486 |
+
style={{ background: "#222" }}
|
| 487 |
+
/>
|
| 488 |
+
<canvas
|
| 489 |
+
ref={canvasRef}
|
| 490 |
+
className="absolute top-0 left-0 w-full h-full pointer-events-none"
|
| 491 |
+
style={{ zIndex: 10, pointerEvents: "none" }}
|
| 492 |
+
/>
|
| 493 |
+
<button
|
| 494 |
+
className="mt-4 px-6 py-2 rounded-lg bg-blue-600 text-white font-semibold"
|
| 495 |
+
onClick={handleToggleExampleProcessing}
|
| 496 |
+
>
|
| 497 |
+
{exampleProcessing ? "Stop Processing" : "Start Processing"}
|
| 498 |
+
</button>
|
| 499 |
+
</div>
|
| 500 |
+
)}
|
| 501 |
+
{processing && <div className="text-blue-400 mt-2">Processing frame...</div>}
|
| 502 |
+
{error && <div className="text-red-400 mt-2">Error: {error}</div>}
|
| 503 |
+
<button
|
| 504 |
+
className="mt-4 px-6 py-2 rounded-lg bg-gray-600 text-white font-semibold"
|
| 505 |
+
onClick={handleTestDrawBox}
|
| 506 |
+
>
|
| 507 |
+
Test Draw Box
|
| 508 |
+
</button>
|
| 509 |
+
<div className="mt-2 p-2 bg-gray-800 rounded text-xs">
|
| 510 |
+
<div>Canvas: {canvasDims ? `${canvasDims.w}x${canvasDims.h}` : "-"} | Video: {videoDims ? `${videoDims.w}x${videoDims.h}` : "-"}</div>
|
| 511 |
+
<div>Raw Model Output:</div>
|
| 512 |
+
<pre className="overflow-x-auto max-h-32 whitespace-pre-wrap">{debugOutput}</pre>
|
| 513 |
+
</div>
|
| 514 |
+
</div>
|
| 515 |
+
)}
|
| 516 |
+
</div>
|
| 517 |
+
</div>
|
| 518 |
+
</div>
|
| 519 |
+
);
|
| 520 |
}
|