Dža ko lil

VGGT-1B

Integriruj

Modelosko id vggt-1b
# Run the model; wait up to 20 s for the finished result in the same call
curl -X POST "https://api.doitong.com/v1/run/vggt-1b?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{}'

# Still "queued" or "running"? Poll with the id from the response:
curl "https://api.doitong.com/v1/requests/REQUEST_ID?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY"

Dži $1.755 pe jekh keripe čargeime palal realno labaripen, o aver si irime Phir les kate

Lel API klidin
Meta Dikhipen

Rekonstruil jekh 3D scena katar dži 48 imidžura vaj jekh video, thaj sikavel depth-mapura, kamera-poseura thaj point-cloud

Meta · Dada (JSON)

Tiro prvo rezultato si kate

Ker tele jekh mangipen thaj phir les tire API klidinosa. E rezultatur si kopirime amare arakhipneste, vično vaj 7 dives bilačhes.

Ker jekh mangipen

O kontrolura kerde si katar e schemaske. O kodo thaj e cena menjinen sar tu len menjines.

Lučhe images · video · alpha_blend_onto · array_encoding · frame_sampling_rate · confidence_percentile · max_points · export_prediction_data · export_depth_maps · export_point_cloud · enable_safety_checker · exclude_keys
image_urls
Čhiv jekh fajl, vaj · dži 48

Reconstruction source kipon, sako padded ande square thaj resized longest dimension-eske 518 pixel-a. JPEG, PNG, WEBP, thaj sako aver kipo formato savo Pillow šaj dekodujel si akceptime.

video_url
Čhiv jekh fajl, vaj

Opcionalno źuravdi video savi le sampled kadrenge palal image_urls ande jekh inputo. Sampling kerel pes sa `frame_sampling_rate`-to kadro, mindig le angluno thaj paluno kadro inklyuzivno.

alpha_blend_onto standardno white

Alpha-channel keripe del štare opcije: 'keep' vazdel o alpha bi te paruvél le originalno pixelengo; 'mean' kompozil pe ImageNet mean RGB. Solidno backgroundoske választin 'white' vaj 'black'.

array_encoding standardno base64

Formato e masivengo arakhipnasko vaš predikciake JSON fajlo, kaj 'list' lel but maj but than sar 'base64'. Alosar 'list' vaš andre-thode JSON liste vaj 'base64' vaš objektura save inćaren {data, shape, dtype}.

standardno 24

Intervalo kadrenge kerdo numa kana video si dinăz `video_url`; averfjal nashti kerel khanchi. Duj granicake kadra astavena pes, thaj sa n-to kadro majralo sample-kerel pes.

standardno 50

Confidence-bazirime filtracijake procento vaš point-cloud keripe: 0 mukel sa; aver vrednosti vazdena avri kad procento pointengo, katar o majcikno confidence.

standardno 250000

GLB point-count granica, standardno 250,000. Te avel majcikno serialization thaj upload latency, jekh ravnomerno subsamplime selekcija lil pes kana o cloud si majbaro kadalestar budžeto.

Kluchi save sa prediction JSON fajlendar.

POST /run/vggt-1b?wait=20 hesab → $1.755
# Run the model; wait up to 20 s for the finished result in the same call
curl -X POST "https://api.doitong.com/v1/run/vggt-1b?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{}'

# Still "queued" or "running"? Poll with the id from the response:
curl "https://api.doitong.com/v1/requests/REQUEST_ID?wait=20" \
  -H "Authorization: Bearer $DOITONG_API_KEY"

Phandel kava mangipen

Tiro kluč ačhel kadale browsereste: bičhalel pes numa ko Doitong API thaj šuvav pes (localStorage) numa te tiknes Remember. Le jekh API ključ

Palal VGGT-1B

Estimirarel geometrija katar but dikhipen. Dža dži ko 48 kipo andro `image_urls`, opcionalno thov `video_url` savo si samplingime sako `frame_sampling_rate` kadro. Del depth mape pala sako kadro, kameraske extrinsics thaj intrinsics, raw prediction fajlo, thaj GLB point cloud e kameraske konenca. `confidence_percentile` thaj `max_points` kontroliren kaboro dense si o point cloud.

Parametra

Bičhalav vaj e kanonikano nav vaj e native nav, na duj. Kanonikane nava

NavTip standardnoMuklo Vakeripe
image_urlsopcionalno images
Tipurl[] · image
≤ 48 artiklura
Reconstruction source kipon, sako padded ande square thaj resized longest dimension-eske 518 pixel-a. JPEG, PNG, WEBP, thaj sako aver kipo formato savo Pillow šaj dekodujel si akceptime.
video_urlopcionalno video
Tipurl · video
Opcionalno źuravdi video savi le sampled kadrenge palal image_urls ande jekh inputo. Sampling kerel pes sa `frame_sampling_rate`-to kadro, mindig le angluno thaj paluno kadro inklyuzivno.
frame_sampling_rateopcionalno
Tipinteger
standardno24
1 – 1,000
Intervalo kadrenge kerdo numa kana video si dinăz `video_url`; averfjal nashti kerel khanchi. Duj granicake kadra astavena pes, thaj sa n-to kadro majralo sample-kerel pes.
alpha_blend_ontoopcionalno
Tipenum
standardnowhite
keepwhiteblackmean
Alpha-channel keripe del štare opcije: 'keep' vazdel o alpha bi te paruvél le originalno pixelengo; 'mean' kompozil pe ImageNet mean RGB. Solidno backgroundoske választin 'white' vaj 'black'.
export_prediction_dataopcionalno
Tipboolean
standardnotrue
Exportisar raw predictiona sar jekhipen JSON fajlo vaš sako kadro. Sako fajlo si depth thaj lesko confidence, masko, thaj encoded camera pose.
export_depth_mapsopcionalno
Tipboolean
standardnotrue
Exportisar depth rastra linearno mappingesa `depth_range` katar pixelengo vrednost 0–65535. Sako kadro lel peske grayscale PNG fajlo 16-bit preciziteta.
export_point_cloudopcionalno
Tipboolean
standardnotrue
Exportisar o fused colored point cloud ande GLB format, estimated kamerenca sar jekhipen cone mesh ande o scena.
array_encodingopcionalno
Tipenum
standardnobase64
base64list
Formato e masivengo arakhipnasko vaš predikciake JSON fajlo, kaj 'list' lel but maj but than sar 'base64'. Alosar 'list' vaš andre-thode JSON liste vaj 'base64' vaš objektura save inćaren {data, shape, dtype}.
exclude_keysopcionalno
Tipenum[]
Kluchi save sa prediction JSON fajlendar.
confidence_percentileopcionalno
Tipnumber
standardno50
0 – 100
Confidence-bazirime filtracijake procento vaš point-cloud keripe: 0 mukel sa; aver vrednosti vazdena avri kad procento pointengo, katar o majcikno confidence.
max_pointsopcionalno
Tipinteger
standardno250000
1,000 – 5,000,000
GLB point-count granica, standardno 250,000. Te avel majcikno serialization thaj upload latency, jekh ravnomerno subsamplime selekcija lil pes kana o cloud si majbaro kadalestar budžeto.
enable_safety_checkeropcionalno
Tipboolean
standardnotrue
Safety-checking togglo vaš inputo kipon thaj video.
trubul Kanonikano nav, vi akceptime

Fajloske parametra lel jekh publikano https:// URL, jekh data: URI dži 10 MB, vaj e file_url uploadoski. Fajlura

Rezultato

GET /requests/req_8k2m4q7x1z9c3v5b6n0p2r4t
{
  "id": "req_8k2m4q7x1z9c3v5b6n0p2r4t",
  "object": "request",
  "model": "vggt-1b",
  "status": "succeeded",
  "created_at": "2026-10-20T12:00:00Z",
  "started_at": "2026-10-20T12:00:02Z",
  "completed_at": "2026-10-20T12:00:09Z",
  "queue_position": null,
  "cost": {
    "credits": 1755,
    "usd": "1.755",
    "charged": 1755,
    "refunded": 0
  },
  "input": {},
  "output": {
    "point_cloud": {
      "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.jpeg",
      "content_type": "image/jpeg",
      "file_name": "0-output.jpeg",
      "file_size": 1843200
    },
    "depth_maps": [
      {
        "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.jpeg",
        "content_type": "image/jpeg",
        "file_name": "0-output.jpeg",
        "file_size": 1843200,
        "width": 1024,
        "height": 768
      }
    ],
    "prediction_data": [
      {
        "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.jpeg",
        "content_type": "image/jpeg",
        "file_name": "0-output.jpeg",
        "file_size": 1843200
      }
    ],
    "num_frames": 1,
    "depth_range": [
      1.0
    ],
    "extrinsics": [
      [
        [
          1.0
        ]
      ]
    ],
    "intrinsics": [
      [
        [
          1.0
        ]
      ]
    ],
    "timings": {}
  },
  "outputs": [
    {
      "type": "file",
      "url": "https://content.doitong.com/dev-api/2026/10/20/req_8k2m4q7x1z9c3v5b6n0p2r4t/0-output.jpeg",
      "content_type": "image/jpeg",
      "file_size": 1843200,
      "stored": true
    }
  ],
  "error": null,
  "metrics": {
    "inference_seconds": 6.8
  },
  "urls": {
    "self": "https://api.doitong.com/v1/requests/req_8k2m4q7x1z9c3v5b6n0p2r4t",
    "cancel": "https://api.doitong.com/v1/requests/req_8k2m4q7x1z9c3v5b6n0p2r4t/cancel"
  }
}
Tire fajlura, tiro alosaripenSako outputs[].url si kopija amare arakhipneste: vično (čargeime per GB-mase) vaj 7 dives bilačhes, tiro kontoskere regulaciatar vaj ?storage=.
Sa jekh forma sako modeloθartype, url, content_type, width, height thaj duration, varesave imajinake, video thaj audio-ke. output si o modelosko korkoro resultosko objekto.
Outputoske fieldura
point_cloud
object {url, content_type, file_name, file_size} | nullPoint-cloud export savo avel kana `export_point_cloud` si true; lesko GLB scena jekh kerel o fused cloud kamerake conenca.
depth_maps
array of object {url, content_type, file_name, file_size, width, height}Eksporto e dungipnaskere mapengo si šajutno kana `export_depth_maps` si true. E elementura džan palal e źutne kadronengi sekvenca, thaj svako kadro si sikavdo sar PNG rastero ande siva nijanse, e 16-bit dungipnasa.
prediction_data
array of object {url, content_type, file_name, file_size}Eksporto e neprocesime predikciako: kana `export_prediction_data` si true, svako kadro si les pesko JSON fajlo, thaj e fajlo džan palal e kadronengi sekvenca.
num_frames
integerTotal rekonstruime kadronengo gin, kipon thaj kadronenca save videoestar.
depth_range
array of numberDepth bounds [min, max] save `depth_maps` quantization. Vares depth sar min + pixel * (max - min) / 65535.
extrinsics
array of array of array of numberCamera-from-world extrinsic matrixa, 3x4, vaš sako kadro; OpenCV axes dikhen right x-este, tele y-este, thaj anglune z-este.
intrinsics
array of array of array of numberPinhole intrinsic matrixa, 3x3, vaš sako kadro, pixel coordinates-e ande o modelesko 518x518 kadro.
timings
objectGelome wall-clock vaxt vaš sako stage, ande sekunda.

AI agentenge: kadija rig sar Markdown si kate https://doitong.com/rom/developers/models/vggt-1b.md. Sa e modelura thaj e gidura: https://doitong.com/developers/llms.txt