ONNX Model Benchmark Report Collection

[HanClinto]

This issue is to collect user contributions of ONNX model benchmarking reports generated on their hardware in order to crowdsource various performance numbers. If you would like to contribute, please run scripts/benchmark_onnx_models.py and submit your results.

[HanClinto]

ONNX CPU benchmark — Apple M1 Pro

• Source: 29ad7ce (dirty) on main
• System: macOS-14.7.1-arm64-arm-64bit · 10P/10L · 32768.0 MiB RAM, 6842.3 MiB available
• Runtime: ONNX Runtime 1.23.2 · Python 3.10.10

Mean/median/P90 are steady-state ms; first run and warmups are excluded.

Model	Version	SHA-256 prefix	Threads	Provider	Mean	Median	P90	First	Size MiB
cornelius	1.164	e190c0f3b782	1	CPUExecutionProvider	48.04	47.80	49.35	55.98	8.23
cornelius	1.164	e190c0f3b782	2	CPUExecutionProvider	27.93	27.54	29.31	34.08	8.23
cornelius	1.164	e190c0f3b782	4	CPUExecutionProvider	21.68	21.36	24.31	23.50	8.23
milo	1.0.0	bd13d8d60383	1	CPUExecutionProvider	71.74	71.20	73.39	78.86	4.95
milo	1.0.0	bd13d8d60383	2	CPUExecutionProvider	46.25	46.95	48.30	49.83	4.95
milo	1.0.0	bd13d8d60383	4	CPUExecutionProvider	32.01	31.63	32.48	35.65	4.95

Full JSON report, source metadata, and model metadata (build/model_benchmarks/onnx-cpu-20260428T233444Z.json)

{
  "created_at": "2026-04-28T23:34:44.064469+00:00",
  "source": {
    "git_commit": "29ad7ce20b77f28b8890663393c4ffad764c801e",
    "git_commit_short": "29ad7ce",
    "git_branch": "main",
    "git_commit_date": "2026-04-28T18:54:13-04:00",
    "git_remote_url": "git@github.com:HanClinto/CollectorVision.git",
    "git_dirty": true,
    "git_status_porcelain": [
      "?? docs/edge_model_optimization.md",
      "?? examples/web_scanner/model_benchmark.html",
      "?? scripts/benchmark_onnx_models.py"
    ]
  },
  "system": {
    "platform": "macOS-14.7.1-arm64-arm-64bit",
    "system": "Darwin",
    "release": "23.6.0",
    "machine": "arm64",
    "processor": "arm",
    "cpu_model": "Apple M1 Pro",
    "cpu_logical_cores": 10,
    "cpu_physical_cores": 10,
    "memory_total_mib": 32768.0,
    "memory_available_mib": 6842.3,
    "python": "3.10.10",
    "onnxruntime": "1.23.2",
    "available_providers": [
      "CoreMLExecutionProvider",
      "AzureExecutionProvider",
      "CPUExecutionProvider"
    ]
  },
  "benchmarks": [
    {
      "name": "cornelius",
      "path": "collector_vision/weights/cornelius.onnx",
      "size_mib": 8.2252,
      "model": {
        "version": "1.164",
        "codename": "cornelius",
        "task": "card-corner-detection",
        "architecture": "mobilevit-xxs-simcc",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "164",
        "input_size": "384",
        "sha256": "sha256:e190c0f3b7829137c64eaa2ea1f0e7c802e2780d51d0325b88c5a006e24578a6",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.164",
          "task": "card-corner-detection",
          "codename": "cornelius",
          "test_iou": "0.7234",
          "base_model": "apple/mobilevit-xx-small",
          "input_size": "384",
          "architecture": "mobilevit-xxs-simcc",
          "training_epochs": "164"
        }
      },
      "provider": "CPUExecutionProvider",
      "threads": 1,
      "input_shape": [
        1,
        3,
        384,
        384
      ],
      "outputs": [
        "corners",
        "presence",
        "sharpness"
      ],
      "session_create_ms": 73.3271,
      "first_run_ms": 55.9777,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 48.038,
        "median": 47.7971,
        "p90": 49.3469,
        "p95": 49.8207,
        "min": 46.8362,
        "max": 51.4135
      },
      "profile": null
    },
    {
      "name": "cornelius",
      "path": "collector_vision/weights/cornelius.onnx",
      "size_mib": 8.2252,
      "model": {
        "version": "1.164",
        "codename": "cornelius",
        "task": "card-corner-detection",
        "architecture": "mobilevit-xxs-simcc",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "164",
        "input_size": "384",
        "sha256": "sha256:e190c0f3b7829137c64eaa2ea1f0e7c802e2780d51d0325b88c5a006e24578a6",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.164",
          "task": "card-corner-detection",
          "codename": "cornelius",
          "test_iou": "0.7234",
          "base_model": "apple/mobilevit-xx-small",
          "input_size": "384",
          "architecture": "mobilevit-xxs-simcc",
          "training_epochs": "164"
        }
      },
      "provider": "CPUExecutionProvider",
      "threads": 2,
      "input_shape": [
        1,
        3,
        384,
        384
      ],
      "outputs": [
        "corners",
        "presence",
        "sharpness"
      ],
      "session_create_ms": 30.5438,
      "first_run_ms": 34.0811,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 27.9338,
        "median": 27.5358,
        "p90": 29.3078,
        "p95": 29.683,
        "min": 27.1716,
        "max": 33.2912
      },
      "profile": null
    },
    {
      "name": "cornelius",
      "path": "collector_vision/weights/cornelius.onnx",
      "size_mib": 8.2252,
      "model": {
        "version": "1.164",
        "codename": "cornelius",
        "task": "card-corner-detection",
        "architecture": "mobilevit-xxs-simcc",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "164",
        "input_size": "384",
        "sha256": "sha256:e190c0f3b7829137c64eaa2ea1f0e7c802e2780d51d0325b88c5a006e24578a6",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.164",
          "task": "card-corner-detection",
          "codename": "cornelius",
          "test_iou": "0.7234",
          "base_model": "apple/mobilevit-xx-small",
          "input_size": "384",
          "architecture": "mobilevit-xxs-simcc",
          "training_epochs": "164"
        }
      },
      "provider": "CPUExecutionProvider",
      "threads": 4,
      "input_shape": [
        1,
        3,
        384,
        384
      ],
      "outputs": [
        "corners",
        "presence",
        "sharpness"
      ],
      "session_create_ms": 33.5181,
      "first_run_ms": 23.4972,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 21.6846,
        "median": 21.3614,
        "p90": 24.3149,
        "p95": 24.5826,
        "min": 19.2036,
        "max": 25.6825
      },
      "profile": null
    },
    {
      "name": "milo",
      "path": "collector_vision/weights/milo.onnx",
      "size_mib": 4.9506,
      "model": {
        "version": "1.0.0",
        "codename": "milo",
        "task": "card-embedding",
        "architecture": "mobilevit-xxs-arcface",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "15",
        "input_size": "448",
        "sha256": "sha256:bd13d8d60383c69da04dce261f32e93fdaeaa8fd618fbc991e7385f71b3d45df",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.0.0",
          "task": "card-embedding",
          "codename": "milo",
          "base_model": "apple/mobilevit-xx-small",
          "architecture": "mobilevit-xxs-arcface",
          "input_size": "448",
          "training_epochs": "15",
          "embedding_dim": "128"
        }
      },
      "provider": "CPUExecutionProvider",
      "threads": 1,
      "input_shape": [
        1,
        3,
        448,
        448
      ],
      "outputs": [
        "embedding"
      ],
      "session_create_ms": 27.0908,
      "first_run_ms": 78.8567,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 71.7397,
        "median": 71.2007,
        "p90": 73.3859,
        "p95": 73.877,
        "min": 70.5843,
        "max": 75.2991
      },
      "profile": null
    },
    {
      "name": "milo",
      "path": "collector_vision/weights/milo.onnx",
      "size_mib": 4.9506,
      "model": {
        "version": "1.0.0",
        "codename": "milo",
        "task": "card-embedding",
        "architecture": "mobilevit-xxs-arcface",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "15",
        "input_size": "448",
        "sha256": "sha256:bd13d8d60383c69da04dce261f32e93fdaeaa8fd618fbc991e7385f71b3d45df",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.0.0",
          "task": "card-embedding",
          "codename": "milo",
          "base_model": "apple/mobilevit-xx-small",
          "architecture": "mobilevit-xxs-arcface",
          "input_size": "448",
          "training_epochs": "15",
          "embedding_dim": "128"
        }
      },
      "provider": "CPUExecutionProvider",
      "threads": 2,
      "input_shape": [
        1,
        3,
        448,
        448
      ],
      "outputs": [
        "embedding"
      ],
      "session_create_ms": 25.6949,
      "first_run_ms": 49.8327,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 46.2512,
        "median": 46.9548,
        "p90": 48.2977,
        "p95": 48.5399,
        "min": 42.2865,
        "max": 51.3959
      },
      "profile": null
    },
    {
      "name": "milo",
      "path": "collector_vision/weights/milo.onnx",
      "size_mib": 4.9506,
      "model": {
        "version": "1.0.0",
        "codename": "milo",
        "task": "card-embedding",
        "architecture": "mobilevit-xxs-arcface",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "15",
        "input_size": "448",
        "sha256": "sha256:bd13d8d60383c69da04dce261f32e93fdaeaa8fd618fbc991e7385f71b3d45df",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.0.0",
          "task": "card-embedding",
          "codename": "milo",
          "base_model": "apple/mobilevit-xx-small",
          "architecture": "mobilevit-xxs-arcface",
          "input_size": "448",
          "training_epochs": "15",
          "embedding_dim": "128"
        }
      },
      "provider": "CPUExecutionProvider",
      "threads": 4,
      "input_shape": [
        1,
        3,
        448,
        448
      ],
      "outputs": [
        "embedding"
      ],
      "session_create_ms": 25.942,
      "first_run_ms": 35.646,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 32.006,
        "median": 31.6273,
        "p90": 32.4844,
        "p95": 33.8295,
        "min": 30.8665,
        "max": 43.6581
      },
      "profile": null
    }
  ]
}

[HanClinto]

ONNX CPU benchmark — Apple M1 Pro

• Source: 29ad7ce (dirty) on main
• System: macOS-14.7.1-arm64-arm-64bit · 10P/10L · 32768.0 MiB RAM, 7124.9 MiB available
• Runtime: ONNX Runtime 1.23.2 · Python 3.10.10

Mean/median/P90 are steady-state ms; first run and warmups are excluded.

Model	Version	Threads	Provider	Mean ms	Median ms	P90 ms	First ms	Size MiB
cornelius	1.164	1	CoreMLExecutionProvider	46.45	45.26	52.42	106.99	8.23
cornelius	1.164	2	CoreMLExecutionProvider	56.20	55.22	59.90	86.24	8.23
cornelius	1.164	4	CoreMLExecutionProvider	54.22	52.74	59.09	74.06	8.23
cornelius	1.164	1	CPUExecutionProvider	48.51	48.07	50.14	51.82	8.23
cornelius	1.164	2	CPUExecutionProvider	29.43	29.20	31.76	33.22	8.23
cornelius	1.164	4	CPUExecutionProvider	21.98	22.05	22.40	24.40	8.23
milo	1.0.0	1	CoreMLExecutionProvider	70.47	69.32	74.48	142.26	4.95
milo	1.0.0	2	CoreMLExecutionProvider	66.15	65.75	68.32	105.96	4.95
milo	1.0.0	4	CoreMLExecutionProvider	61.00	60.72	62.92	97.39	4.95
milo	1.0.0	1	CPUExecutionProvider	72.75	72.57	73.77	76.70	4.95
milo	1.0.0	2	CPUExecutionProvider	48.94	49.49	50.08	52.71	4.95
milo	1.0.0	4	CPUExecutionProvider	33.18	33.22	33.55	34.82	4.95

Full JSON report (build/model_benchmarks/onnx-cpu-20260428T234348Z.json)

{
  "created_at": "2026-04-28T23:43:48.867032+00:00",
  "source": {
    "git_commit": "29ad7ce20b77f28b8890663393c4ffad764c801e",
    "git_commit_short": "29ad7ce",
    "git_branch": "main",
    "git_commit_date": "2026-04-28T18:54:13-04:00",
    "git_remote_url": "git@github.com:HanClinto/CollectorVision.git",
    "git_dirty": true,
    "git_status_porcelain": [
      "?? docs/edge_model_optimization.md",
      "?? examples/web_scanner/model_benchmark.html",
      "?? scripts/benchmark_onnx_models.py"
    ]
  },
  "system": {
    "platform": "macOS-14.7.1-arm64-arm-64bit",
    "system": "Darwin",
    "release": "23.6.0",
    "machine": "arm64",
    "processor": "arm",
    "cpu_model": "Apple M1 Pro",
    "cpu_logical_cores": 10,
    "cpu_physical_cores": 10,
    "memory_total_mib": 32768.0,
    "memory_available_mib": 7124.9,
    "python": "3.10.10",
    "onnxruntime": "1.23.2",
    "available_providers": [
      "CoreMLExecutionProvider",
      "AzureExecutionProvider",
      "CPUExecutionProvider"
    ]
  },
  "benchmarks": [
    {
      "name": "cornelius",
      "path": "collector_vision/weights/cornelius.onnx",
      "size_mib": 8.2252,
      "model": {
        "version": "1.164",
        "codename": "cornelius",
        "task": "card-corner-detection",
        "architecture": "mobilevit-xxs-simcc",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "164",
        "input_size": "384",
        "sha256": "sha256:e190c0f3b7829137c64eaa2ea1f0e7c802e2780d51d0325b88c5a006e24578a6",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.164",
          "task": "card-corner-detection",
          "codename": "cornelius",
          "test_iou": "0.7234",
          "base_model": "apple/mobilevit-xx-small",
          "input_size": "384",
          "architecture": "mobilevit-xxs-simcc",
          "training_epochs": "164"
        }
      },
      "provider": "CoreMLExecutionProvider",
      "provider_chain": [
        "CoreMLExecutionProvider",
        "CPUExecutionProvider"
      ],
      "session_providers": [
        "CoreMLExecutionProvider",
        "CPUExecutionProvider"
      ],
      "threads": 1,
      "input_shape": [
        1,
        3,
        384,
        384
      ],
      "outputs": [
        "corners",
        "presence",
        "sharpness"
      ],
      "session_create_ms": 3211.6424,
      "first_run_ms": 106.9905,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 46.4494,
        "median": 45.2641,
        "p90": 52.4248,
        "p95": 53.8118,
        "min": 43.0353,
        "max": 54.7805
      },
      "profile": null
    },
    {
      "name": "cornelius",
      "path": "collector_vision/weights/cornelius.onnx",
      "size_mib": 8.2252,
      "model": {
        "version": "1.164",
        "codename": "cornelius",
        "task": "card-corner-detection",
        "architecture": "mobilevit-xxs-simcc",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "164",
        "input_size": "384",
        "sha256": "sha256:e190c0f3b7829137c64eaa2ea1f0e7c802e2780d51d0325b88c5a006e24578a6",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.164",
          "task": "card-corner-detection",
          "codename": "cornelius",
          "test_iou": "0.7234",
          "base_model": "apple/mobilevit-xx-small",
          "input_size": "384",
          "architecture": "mobilevit-xxs-simcc",
          "training_epochs": "164"
        }
      },
      "provider": "CoreMLExecutionProvider",
      "provider_chain": [
        "CoreMLExecutionProvider",
        "CPUExecutionProvider"
      ],
      "session_providers": [
        "CoreMLExecutionProvider",
        "CPUExecutionProvider"
      ],
      "threads": 2,
      "input_shape": [
        1,
        3,
        384,
        384
      ],
      "outputs": [
        "corners",
        "presence",
        "sharpness"
      ],
      "session_create_ms": 3026.456,
      "first_run_ms": 86.2363,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 56.2024,
        "median": 55.2195,
        "p90": 59.9017,
        "p95": 63.8792,
        "min": 51.7642,
        "max": 74.7153
      },
      "profile": null
    },
    {
      "name": "cornelius",
      "path": "collector_vision/weights/cornelius.onnx",
      "size_mib": 8.2252,
      "model": {
        "version": "1.164",
        "codename": "cornelius",
        "task": "card-corner-detection",
        "architecture": "mobilevit-xxs-simcc",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "164",
        "input_size": "384",
        "sha256": "sha256:e190c0f3b7829137c64eaa2ea1f0e7c802e2780d51d0325b88c5a006e24578a6",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.164",
          "task": "card-corner-detection",
          "codename": "cornelius",
          "test_iou": "0.7234",
          "base_model": "apple/mobilevit-xx-small",
          "input_size": "384",
          "architecture": "mobilevit-xxs-simcc",
          "training_epochs": "164"
        }
      },
      "provider": "CoreMLExecutionProvider",
      "provider_chain": [
        "CoreMLExecutionProvider",
        "CPUExecutionProvider"
      ],
      "session_providers": [
        "CoreMLExecutionProvider",
        "CPUExecutionProvider"
      ],
      "threads": 4,
      "input_shape": [
        1,
        3,
        384,
        384
      ],
      "outputs": [
        "corners",
        "presence",
        "sharpness"
      ],
      "session_create_ms": 3135.7175,
      "first_run_ms": 74.0605,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 54.2163,
        "median": 52.7384,
        "p90": 59.0853,
        "p95": 62.1329,
        "min": 50.8297,
        "max": 66.187
      },
      "profile": null
    },
    {
      "name": "cornelius",
      "path": "collector_vision/weights/cornelius.onnx",
      "size_mib": 8.2252,
      "model": {
        "version": "1.164",
        "codename": "cornelius",
        "task": "card-corner-detection",
        "architecture": "mobilevit-xxs-simcc",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "164",
        "input_size": "384",
        "sha256": "sha256:e190c0f3b7829137c64eaa2ea1f0e7c802e2780d51d0325b88c5a006e24578a6",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.164",
          "task": "card-corner-detection",
          "codename": "cornelius",
          "test_iou": "0.7234",
          "base_model": "apple/mobilevit-xx-small",
          "input_size": "384",
          "architecture": "mobilevit-xxs-simcc",
          "training_epochs": "164"
        }
      },
      "provider": "CPUExecutionProvider",
      "provider_chain": [
        "CPUExecutionProvider"
      ],
      "session_providers": [
        "CPUExecutionProvider"
      ],
      "threads": 1,
      "input_shape": [
        1,
        3,
        384,
        384
      ],
      "outputs": [
        "corners",
        "presence",
        "sharpness"
      ],
      "session_create_ms": 36.037,
      "first_run_ms": 51.8173,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 48.5062,
        "median": 48.072,
        "p90": 50.1421,
        "p95": 50.3319,
        "min": 47.156,
        "max": 50.6798
      },
      "profile": null
    },
    {
      "name": "cornelius",
      "path": "collector_vision/weights/cornelius.onnx",
      "size_mib": 8.2252,
      "model": {
        "version": "1.164",
        "codename": "cornelius",
        "task": "card-corner-detection",
        "architecture": "mobilevit-xxs-simcc",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "164",
        "input_size": "384",
        "sha256": "sha256:e190c0f3b7829137c64eaa2ea1f0e7c802e2780d51d0325b88c5a006e24578a6",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.164",
          "task": "card-corner-detection",
          "codename": "cornelius",
          "test_iou": "0.7234",
          "base_model": "apple/mobilevit-xx-small",
          "input_size": "384",
          "architecture": "mobilevit-xxs-simcc",
          "training_epochs": "164"
        }
      },
      "provider": "CPUExecutionProvider",
      "provider_chain": [
        "CPUExecutionProvider"
      ],
      "session_providers": [
        "CPUExecutionProvider"
      ],
      "threads": 2,
      "input_shape": [
        1,
        3,
        384,
        384
      ],
      "outputs": [
        "corners",
        "presence",
        "sharpness"
      ],
      "session_create_ms": 32.3772,
      "first_run_ms": 33.2202,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 29.429,
        "median": 29.2019,
        "p90": 31.7618,
        "p95": 32.6466,
        "min": 27.3599,
        "max": 33.3587
      },
      "profile": null
    },
    {
      "name": "cornelius",
      "path": "collector_vision/weights/cornelius.onnx",
      "size_mib": 8.2252,
      "model": {
        "version": "1.164",
        "codename": "cornelius",
        "task": "card-corner-detection",
        "architecture": "mobilevit-xxs-simcc",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "164",
        "input_size": "384",
        "sha256": "sha256:e190c0f3b7829137c64eaa2ea1f0e7c802e2780d51d0325b88c5a006e24578a6",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.164",
          "task": "card-corner-detection",
          "codename": "cornelius",
          "test_iou": "0.7234",
          "base_model": "apple/mobilevit-xx-small",
          "input_size": "384",
          "architecture": "mobilevit-xxs-simcc",
          "training_epochs": "164"
        }
      },
      "provider": "CPUExecutionProvider",
      "provider_chain": [
        "CPUExecutionProvider"
      ],
      "session_providers": [
        "CPUExecutionProvider"
      ],
      "threads": 4,
      "input_shape": [
        1,
        3,
        384,
        384
      ],
      "outputs": [
        "corners",
        "presence",
        "sharpness"
      ],
      "session_create_ms": 30.0212,
      "first_run_ms": 24.3981,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 21.9824,
        "median": 22.0542,
        "p90": 22.4015,
        "p95": 22.5224,
        "min": 21.1407,
        "max": 22.605
      },
      "profile": null
    },
    {
      "name": "milo",
      "path": "collector_vision/weights/milo.onnx",
      "size_mib": 4.9506,
      "model": {
        "version": "1.0.0",
        "codename": "milo",
        "task": "card-embedding",
        "architecture": "mobilevit-xxs-arcface",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "15",
        "input_size": "448",
        "sha256": "sha256:bd13d8d60383c69da04dce261f32e93fdaeaa8fd618fbc991e7385f71b3d45df",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.0.0",
          "task": "card-embedding",
          "codename": "milo",
          "base_model": "apple/mobilevit-xx-small",
          "architecture": "mobilevit-xxs-arcface",
          "input_size": "448",
          "training_epochs": "15",
          "embedding_dim": "128"
        }
      },
      "provider": "CoreMLExecutionProvider",
      "provider_chain": [
        "CoreMLExecutionProvider",
        "CPUExecutionProvider"
      ],
      "session_providers": [
        "CoreMLExecutionProvider",
        "CPUExecutionProvider"
      ],
      "threads": 1,
      "input_shape": [
        1,
        3,
        448,
        448
      ],
      "outputs": [
        "embedding"
      ],
      "session_create_ms": 2893.3792,
      "first_run_ms": 142.2562,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 70.4715,
        "median": 69.3249,
        "p90": 74.4836,
        "p95": 77.8496,
        "min": 67.12,
        "max": 82.3837
      },
      "profile": null
    },
    {
      "name": "milo",
      "path": "collector_vision/weights/milo.onnx",
      "size_mib": 4.9506,
      "model": {
        "version": "1.0.0",
        "codename": "milo",
        "task": "card-embedding",
        "architecture": "mobilevit-xxs-arcface",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "15",
        "input_size": "448",
        "sha256": "sha256:bd13d8d60383c69da04dce261f32e93fdaeaa8fd618fbc991e7385f71b3d45df",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.0.0",
          "task": "card-embedding",
          "codename": "milo",
          "base_model": "apple/mobilevit-xx-small",
          "architecture": "mobilevit-xxs-arcface",
          "input_size": "448",
          "training_epochs": "15",
          "embedding_dim": "128"
        }
      },
      "provider": "CoreMLExecutionProvider",
      "provider_chain": [
        "CoreMLExecutionProvider",
        "CPUExecutionProvider"
      ],
      "session_providers": [
        "CoreMLExecutionProvider",
        "CPUExecutionProvider"
      ],
      "threads": 2,
      "input_shape": [
        1,
        3,
        448,
        448
      ],
      "outputs": [
        "embedding"
      ],
      "session_create_ms": 2880.8976,
      "first_run_ms": 105.9573,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 66.149,
        "median": 65.7549,
        "p90": 68.3197,
        "p95": 70.1239,
        "min": 63.5109,
        "max": 73.6414
      },
      "profile": null
    },
    {
      "name": "milo",
      "path": "collector_vision/weights/milo.onnx",
      "size_mib": 4.9506,
      "model": {
        "version": "1.0.0",
        "codename": "milo",
        "task": "card-embedding",
        "architecture": "mobilevit-xxs-arcface",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "15",
        "input_size": "448",
        "sha256": "sha256:bd13d8d60383c69da04dce261f32e93fdaeaa8fd618fbc991e7385f71b3d45df",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.0.0",
          "task": "card-embedding",
          "codename": "milo",
          "base_model": "apple/mobilevit-xx-small",
          "architecture": "mobilevit-xxs-arcface",
          "input_size": "448",
          "training_epochs": "15",
          "embedding_dim": "128"
        }
      },
      "provider": "CoreMLExecutionProvider",
      "provider_chain": [
        "CoreMLExecutionProvider",
        "CPUExecutionProvider"
      ],
      "session_providers": [
        "CoreMLExecutionProvider",
        "CPUExecutionProvider"
      ],
      "threads": 4,
      "input_shape": [
        1,
        3,
        448,
        448
      ],
      "outputs": [
        "embedding"
      ],
      "session_create_ms": 2804.9676,
      "first_run_ms": 97.3944,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 61.0032,
        "median": 60.7152,
        "p90": 62.9227,
        "p95": 63.0671,
        "min": 58.7337,
        "max": 65.6361
      },
      "profile": null
    },
    {
      "name": "milo",
      "path": "collector_vision/weights/milo.onnx",
      "size_mib": 4.9506,
      "model": {
        "version": "1.0.0",
        "codename": "milo",
        "task": "card-embedding",
        "architecture": "mobilevit-xxs-arcface",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "15",
        "input_size": "448",
        "sha256": "sha256:bd13d8d60383c69da04dce261f32e93fdaeaa8fd618fbc991e7385f71b3d45df",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.0.0",
          "task": "card-embedding",
          "codename": "milo",
          "base_model": "apple/mobilevit-xx-small",
          "architecture": "mobilevit-xxs-arcface",
          "input_size": "448",
          "training_epochs": "15",
          "embedding_dim": "128"
        }
      },
      "provider": "CPUExecutionProvider",
      "provider_chain": [
        "CPUExecutionProvider"
      ],
      "session_providers": [
        "CPUExecutionProvider"
      ],
      "threads": 1,
      "input_shape": [
        1,
        3,
        448,
        448
      ],
      "outputs": [
        "embedding"
      ],
      "session_create_ms": 26.1155,
      "first_run_ms": 76.6961,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 72.7546,
        "median": 72.5699,
        "p90": 73.7656,
        "p95": 74.4312,
        "min": 71.7689,
        "max": 76.8072
      },
      "profile": null
    },
    {
      "name": "milo",
      "path": "collector_vision/weights/milo.onnx",
      "size_mib": 4.9506,
      "model": {
        "version": "1.0.0",
        "codename": "milo",
        "task": "card-embedding",
        "architecture": "mobilevit-xxs-arcface",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "15",
        "input_size": "448",
        "sha256": "sha256:bd13d8d60383c69da04dce261f32e93fdaeaa8fd618fbc991e7385f71b3d45df",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.0.0",
          "task": "card-embedding",
          "codename": "milo",
          "base_model": "apple/mobilevit-xx-small",
          "architecture": "mobilevit-xxs-arcface",
          "input_size": "448",
          "training_epochs": "15",
          "embedding_dim": "128"
        }
      },
      "provider": "CPUExecutionProvider",
      "provider_chain": [
        "CPUExecutionProvider"
      ],
      "session_providers": [
        "CPUExecutionProvider"
      ],
      "threads": 2,
      "input_shape": [
        1,
        3,
        448,
        448
      ],
      "outputs": [
        "embedding"
      ],
      "session_create_ms": 25.7151,
      "first_run_ms": 52.7107,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 48.9422,
        "median": 49.4947,
        "p90": 50.0811,
        "p95": 50.9662,
        "min": 41.1161,
        "max": 51.3806
      },
      "profile": null
    },
    {
      "name": "milo",
      "path": "collector_vision/weights/milo.onnx",
      "size_mib": 4.9506,
      "model": {
        "version": "1.0.0",
        "codename": "milo",
        "task": "card-embedding",
        "architecture": "mobilevit-xxs-arcface",
        "base_model": "apple/mobilevit-xx-small",
        "training_epochs": "15",
        "input_size": "448",
        "sha256": "sha256:bd13d8d60383c69da04dce261f32e93fdaeaa8fd618fbc991e7385f71b3d45df",
        "onnx_producer_name": "pytorch",
        "onnx_graph_name": "main_graph",
        "onnx_domain": "",
        "onnx_model_version": 9223372036854775807,
        "custom_metadata": {
          "version": "1.0.0",
          "task": "card-embedding",
          "codename": "milo",
          "base_model": "apple/mobilevit-xx-small",
          "architecture": "mobilevit-xxs-arcface",
          "input_size": "448",
          "training_epochs": "15",
          "embedding_dim": "128"
        }
      },
      "provider": "CPUExecutionProvider",
      "provider_chain": [
        "CPUExecutionProvider"
      ],
      "session_providers": [
        "CPUExecutionProvider"
      ],
      "threads": 4,
      "input_shape": [
        1,
        3,
        448,
        448
      ],
      "outputs": [
        "embedding"
      ],
      "session_create_ms": 24.8674,
      "first_run_ms": 34.8159,
      "latency_ms": {
        "runs": 50,
        "warmup": 10,
        "note": "first_run_ms and warmup runs are excluded from these steady-state timings",
        "mean": 33.1825,
        "median": 33.2177,
        "p90": 33.547,
        "p95": 33.6258,
        "min": 32.311,
        "max": 34.2487
      },
      "profile": null
    }
  ]
}

[HanClinto]

ONNX browser benchmark

• Source: ccc3d6b
• System: Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:149.0) Gecko/20100101 Firefox/149.0 · 10 logical threads
• Runtime: ONNX Runtime Web · wasm · WASM thread sweep 1, 4
• Isolation: cross-origin isolated true

Mean/median/P90 are steady-state ms; first run and warmups are excluded.

Model	Version	Threads	Provider	Mean ms	Median ms	P90 ms	First ms	Size MiB
cornelius	unknown	1	wasm	94.01	93.01	94.40	114.64	6.49
milo	unknown	1	wasm	144.32	141.79	143.49	147.20	3.58
cornelius	unknown	4	wasm	93.66	93.55	94.99	94.36	6.49
milo	unknown	4	wasm	144.81	141.43	151.78	148.18	3.58

Full JSON report

{
  "createdAt": "2026-04-29T02:03:11.583Z",
  "source": {
    "buildId": "ccc3d6b"
  },
  "userAgent": "Mozilla/5.0 (Macintosh; Intel Mac OS X 10.15; rv:149.0) Gecko/20100101 Firefox/149.0",
  "crossOriginIsolated": true,
  "sharedArrayBuffer": true,
  "hardwareConcurrency": 10,
  "benchmarkMode": "wasm-thread-sweep",
  "backend": "wasm",
  "requestedWasmThreads": 4,
  "wasmThreads": [
    1,
    4
  ],
  "wasmThreadSweep": [
    1,
    4
  ],
  "warmupRuns": 10,
  "measuredRuns": 50,
  "benchmarks": [
    {
      "name": "cornelius",
      "path": "./assets/models/cornelius.onnx",
      "version": "unknown",
      "sizeMiB": 6.49,
      "backend": "wasm",
      "provider": "wasm",
      "threads": 1,
      "inputShape": [
        1,
        3,
        384,
        384
      ],
      "outputNames": [
        "corners",
        "presence",
        "sharpness"
      ],
      "sessionCreateMs": 3726.7,
      "firstRunMs": 114.64,
      "latencyMs": {
        "runs": 50,
        "mean": 94.01,
        "median": 93.01,
        "p90": 94.4,
        "p95": 94.741,
        "min": 91.94,
        "max": 136.9
      }
    },
    {
      "name": "milo",
      "path": "./assets/models/milo.onnx",
      "version": "unknown",
      "sizeMiB": 3.58,
      "backend": "wasm",
      "provider": "wasm",
      "threads": 1,
      "inputShape": [
        1,
        3,
        448,
        448
      ],
      "outputNames": [
        "embedding"
      ],
      "sessionCreateMs": 556.2,
      "firstRunMs": 147.2,
      "latencyMs": {
        "runs": 50,
        "mean": 144.318,
        "median": 141.79,
        "p90": 143.488,
        "p95": 144.228,
        "min": 140.4,
        "max": 231.02
      }
    },
    {
      "name": "cornelius",
      "path": "./assets/models/cornelius.onnx",
      "version": "unknown",
      "sizeMiB": 6.49,
      "backend": "wasm",
      "provider": "wasm",
      "threads": 4,
      "inputShape": [
        1,
        3,
        384,
        384
      ],
      "outputNames": [
        "corners",
        "presence",
        "sharpness"
      ],
      "sessionCreateMs": 139.3,
      "firstRunMs": 94.36,
      "latencyMs": {
        "runs": 50,
        "mean": 93.664,
        "median": 93.55,
        "p90": 94.986,
        "p95": 95.435,
        "min": 92.14,
        "max": 95.9
      }
    },
    {
      "name": "milo",
      "path": "./assets/models/milo.onnx",
      "version": "unknown",
      "sizeMiB": 3.58,
      "backend": "wasm",
      "provider": "wasm",
      "threads": 4,
      "inputShape": [
        1,
        3,
        448,
        448
      ],
      "outputNames": [
        "embedding"
      ],
      "sessionCreateMs": 102.78,
      "firstRunMs": 148.18,
      "latencyMs": {
        "runs": 50,
        "mean": 144.806,
        "median": 141.43,
        "p90": 151.78,
        "p95": 153.775,
        "min": 135.34,
        "max": 233.68
      }
    }
  ]
}

[niothiel]

ONNX browser benchmark

• Source: 9d45a37
• System: Mozilla/5.0 (X11; Linux x86_64; rv:140.0) Gecko/20100101 Firefox/140.0 · 12 logical threads
• Runtime: ONNX Runtime Web · wasm · WASM thread sweep 1, 4
• Isolation: cross-origin isolated true

Mean/median/P90 are steady-state ms; first run and warmups are excluded.

Model	Version	Threads	Provider	Mean ms	Median ms	P90 ms	First ms	Size MiB
cornelius	2.12	1	wasm	110.73	110.93	114.33	136.34	3.76
milo	1.0.0	1	wasm	165.82	164.42	169.38	178.96	3.58
cornelius	2.12	4	wasm	106.97	106.62	107.62	108.36	3.76
milo	1.0.0	4	wasm	165.75	165.23	169.58	164.96	3.58

Full JSON report

{
  "createdAt": "2026-07-14T12:20:47.334Z",
  "source": {
    "buildId": "9d45a37"
  },
  "userAgent": "Mozilla/5.0 (X11; Linux x86_64; rv:140.0) Gecko/20100101 Firefox/140.0",
  "crossOriginIsolated": true,
  "sharedArrayBuffer": true,
  "hardwareConcurrency": 12,
  "benchmarkMode": "wasm-thread-sweep",
  "backend": "wasm",
  "requestedWasmThreads": 4,
  "wasmThreads": [
    1,
    4
  ],
  "wasmThreadSweep": [
    1,
    4
  ],
  "warmupRuns": 10,
  "measuredRuns": 50,
  "benchmarks": [
    {
      "name": "cornelius",
      "path": "./assets/models/detector.onnx",
      "version": "2.12",
      "sizeMiB": 3.76,
      "backend": "wasm",
      "provider": "wasm",
      "threads": 1,
      "inputShape": [
        1,
        3,
        384,
        384
      ],
      "outputNames": [
        "corners",
        "presence",
        "sharpness"
      ],
      "sessionCreateMs": 664.9,
      "firstRunMs": 136.34,
      "latencyMs": {
        "runs": 50,
        "mean": 110.726,
        "median": 110.93,
        "p90": 114.33,
        "p95": 115.873,
        "min": 105.72,
        "max": 117.46
      }
    },
    {
      "name": "milo",
      "path": "./assets/models/milo.onnx",
      "version": "1.0.0",
      "sizeMiB": 3.58,
      "backend": "wasm",
      "provider": "wasm",
      "threads": 1,
      "inputShape": [
        1,
        3,
        448,
        448
      ],
      "outputNames": [
        "embedding"
      ],
      "sessionCreateMs": 407.4,
      "firstRunMs": 178.96,
      "latencyMs": {
        "runs": 50,
        "mean": 165.818,
        "median": 164.42,
        "p90": 169.376,
        "p95": 171.677,
        "min": 162.7,
        "max": 173.12
      }
    },
    {
      "name": "cornelius",
      "path": "./assets/models/detector.onnx",
      "version": "2.12",
      "sizeMiB": 3.76,
      "backend": "wasm",
      "provider": "wasm",
      "threads": 4,
      "inputShape": [
        1,
        3,
        384,
        384
      ],
      "outputNames": [
        "corners",
        "presence",
        "sharpness"
      ],
      "sessionCreateMs": 124.84,
      "firstRunMs": 108.36,
      "latencyMs": {
        "runs": 50,
        "mean": 106.97,
        "median": 106.62,
        "p90": 107.62,
        "p95": 108.477,
        "min": 105.68,
        "max": 115.46
      }
    },
    {
      "name": "milo",
      "path": "./assets/models/milo.onnx",
      "version": "1.0.0",
      "sizeMiB": 3.58,
      "backend": "wasm",
      "provider": "wasm",
      "threads": 4,
      "inputShape": [
        1,
        3,
        448,
        448
      ],
      "outputNames": [
        "embedding"
      ],
      "sessionCreateMs": 96.5,
      "firstRunMs": 164.96,
      "latencyMs": {
        "runs": 50,
        "mean": 165.753,
        "median": 165.23,
        "p90": 169.584,
        "p95": 170.728,
        "min": 162.12,
        "max": 172.38
      }
    }
  ]
}