micro-handpose/test.html at main · svenflow/micro-handpose · GitHub

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
<!DOCTYPE html>
<html lang="en">
<head>
  <meta charset="UTF-8">
  <meta name="viewport" content="width=device-width, initial-scale=1.0">
  <title>micro-handpose test</title>
  <style>
    body { font-family: system-ui; max-width: 600px; margin: 40px auto; padding: 20px; background: #111; color: #eee; }
    pre { background: #222; padding: 15px; border-radius: 8px; overflow-x: auto; font-size: 13px; white-space: pre-wrap; }
    img { max-width: 200px; border-radius: 8px; }
  </style>
</head>
<body>
  <h1>micro-handpose test</h1>
  <img id="testImg" alt="test hand">
  <pre id="log"></pre>

  <script type="module">
    const logEl = document.getElementById('log');
    function log(msg) {
      logEl.textContent += msg + '\n';
      console.log(msg);
    }

    async function run() {
      try {
        log('importing micro-handpose...');
        const { createHandpose } = await import('./dist/index.js');

        log('creating detector...');
        const t0 = performance.now();
        const handpose = await createHandpose({ weightsUrl: '/weights' });
        log(`ready in ${(performance.now() - t0).toFixed(0)}ms`);

        const img = document.getElementById('testImg');
        img.src = '/test_hand.png';
        await new Promise(r => img.onload = r);

        const canvas = new OffscreenCanvas(256, 256);
        const ctx = canvas.getContext('2d');
        ctx.drawImage(img, 0, 0, 256, 256);

        // Correctness
        const result = await handpose.detect(canvas);
        if (result) {
          log(`✅ score=${result.score.toFixed(4)} hand=${result.handedness}`);
        } else {
          log('❌ no hand detected!');
          return;
        }

        // Diagnostic benchmark
        log('\n🔬 Running diagnostic benchmark...\n');
        const diag = await handpose.benchmarkDiagnostic(canvas, 50);

        log('=== TIMING BREAKDOWN (median / min) ===');
        log(`GPU only:              ${diag.gpuOnly.median.toFixed(1)}ms / ${diag.gpuOnly.min.toFixed(1)}ms`);
        log(`mapAsync+wait:         ${diag.mapAsyncOnly.median.toFixed(1)}ms / ${diag.mapAsyncOnly.min.toFixed(1)}ms`);
        log(`mapAsync raw:          ${diag.mapAsyncNoWait.median.toFixed(1)}ms / ${diag.mapAsyncNoWait.min.toFixed(1)}ms`);
        log(`Total e2e:             ${diag.total.median.toFixed(1)}ms / ${diag.total.min.toFixed(1)}ms`);
        log(`Pipelined:             ${diag.pipelined.median.toFixed(1)}ms / ${diag.pipelined.min.toFixed(1)}ms`);
        if (diag.renderReadback) {
          log(`Render readback:       ${diag.renderReadback.median.toFixed(1)}ms / ${diag.renderReadback.min.toFixed(1)}ms`);
        } else {
          log(`Render readback:       N/A (no WebGPU canvas)`);
        }

        log(`\nReadback overhead: ~${(diag.total.median - diag.gpuOnly.median).toFixed(1)}ms`);
        log(`Best approach: ${getBest(diag)}`);

        handpose.dispose();
        log('\ndone.');

      } catch (e) {
        log(`\n❌ error: ${e.message}\n${e.stack}`);
        console.error(e);
      }
    }

    function getBest(diag) {
      const approaches = [
        ['mapAsync', diag.total.median],
        ['pipelined', diag.pipelined.median],
      ];
      if (diag.renderReadback) approaches.push(['render', diag.renderReadback.median]);
      approaches.sort((a, b) => a[1] - b[1]);
      return `${approaches[0][0]} (${approaches[0][1].toFixed(1)}ms)`;
    }

    run();
  </script>
</body>
</html>