[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"project-92722":3},{"id":4,"name":5,"fullName":6,"owner":7,"repo":5,"description":8,"homepage":9,"htmlUrl":9,"language":10,"languages":9,"totalLinesOfCode":9,"stars":11,"forks":12,"watchers":13,"openIssues":14,"contributorsCount":15,"subscribersCount":15,"size":15,"stars1d":15,"stars7d":15,"stars30d":16,"stars90d":15,"forks30d":15,"starsTrendScore":15,"compositeScore":17,"rankGlobal":9,"rankLanguage":9,"license":18,"archived":19,"fork":19,"defaultBranch":20,"hasWiki":19,"hasPages":19,"topics":21,"createdAt":9,"pushedAt":9,"updatedAt":22,"readmeContent":23,"aiSummary":24,"trendingCount":15,"starSnapshotCount":15,"syncStatus":14,"lastSyncTime":25,"discoverSource":26},92722,"SenseNova-Vision","OpenSenseNova\u002FSenseNova-Vision","OpenSenseNova","Vision as Unified Multimodal Generation",null,"Python",86,3,1,2,0,27,44.51,"Apache License 2.0",false,"master",[],"2026-07-22 04:02:06","\u003Cdiv align=\"center\">\n\n# SenseNova-Vision: Vision as Unified Multimodal Generation\n\n\u003C\u002Fdiv>\n\n\u003Cdiv align=\"center\">\n  \u003Ca href=\".\u002FREADME.md\">English\u003C\u002Fa> | \u003Ca href=\".\u002FREADME_CN.md\">简体中文\u003C\u002Fa>\n  \u003Cbr>\n\n  \u003Ca href=\"https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.06560\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FarXiv-SenseNova--Vision-b31b1b.svg\" alt=\"arXiv\">\u003C\u002Fa>\n  \u003Ca href=\".\u002Fdocs\u002FEVAL.md\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FEvaluation-Guide-green\" alt=\"Evaluation Guide\">\u003C\u002Fa>\n  \u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fsensenova\u002FSenseNova-Vision-7B-MoT\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%A4%97%20HuggingFace-Model-yellow\" alt=\"HuggingFace Model\">\u003C\u002Fa>\n  \u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002Fsensenova\u002FSenseNova-Vision-Corpus-50M\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%A4%97%20HuggingFace-Dataset-yellow\" alt=\"HuggingFace Dataset\">\u003C\u002Fa>\n  \u003Ca href=\"https:\u002F\u002Fhuggingface.co\u002Fspaces\u002Fsensenova\u002FSenseNova-Vision\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F%F0%9F%A4%97%20SenseNova--Vision-Demo-Green\" alt=\"SenseNova-Vision Demo\">\u003C\u002Fa>\n  \u003Ca href=\"https:\u002F\u002Fmodelscope.cn\u002Fmodels\u002FSenseNova\u002FSenseNova-Vision-7B-MoT\" target=\"_blank\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002F🤖%20ModelScope-Model-blue\" alt=\"ModelScope Model\">\u003C\u002Fa>\n  \u003Ca href=\".\u002FLICENSE\">\u003Cimg src=\"https:\u002F\u002Fimg.shields.io\u002Fbadge\u002FLicense-Apache%202.0-blue.svg\" alt=\"License\">\u003C\u002Fa>\n\n\u003Cbr>\n  \u003Ca href=\".\u002Fassets\u002Fshowcase\u002Ffig2_one_case_for_all.webp\">\u003Cimg src=\".\u002Fassets\u002Fshowcase\u002Ffig2_one_case_for_all.webp\" alt=\"SenseNova-Vision handles diverse vision tasks in a unified model\" width=\"900\">\u003C\u002Fa>\n\n\u003Cbr>\n  \u003Cimg src=\".\u002Fassets\u002Ffig3_system_overview.webp\" alt=\"SenseNova-Vision system overview\" width=\"900\">\n\u003C\u002Fdiv>\n\n## 📣 Updated News\n\n- `[2026.07.08]` Release of the dataset for [SenseNova-Vision-Corpus-50M](https:\u002F\u002Fhuggingface.co\u002Fdatasets\u002Fsensenova\u002FSenseNova-Vision-Corpus-50M).\n- `[2026.07.08]` Initial release of the weights for [SenseNova-Vision-7B-MoT](https:\u002F\u002Fhuggingface.co\u002Fsensenova\u002FSenseNova-Vision-7B-MoT).\n- `[2026.07.08]` Initial release of the [inference code](https:\u002F\u002Fgithub.com\u002FOpenSenseNova\u002FSenseNova-Vision) for SenseNova-Vision.\n- `[2026.07.08]` Release of the [Technical Report](https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.06560) for SenseNova-Vision.\n\n## 🌟 Overview\n\n🚀 **SenseNova-Vision** formulate computer vision as unified multimodal generation, \nwhere heterogeneous visual tasks are expressed through the native \ntext and image generation spaces of a unified multimodal model (UMM).\nNatural-language instructions and optional visual prompts specify the\ntask, target regions or views, output schema, and decoding convention, while\nthe model responds through native text, image, or mixed text-image generation.\n\nText generation expresses symbolic visual records such as categories, boxes,\npoints, OCR strings, keypoints, and camera parameters. Image generation handles\ndense spatial targets such as segmentation masks, depth maps, surface normals,\nand multi-view point maps. Mixed responses support compositional tasks that\ncombine symbolic and dense outputs. This shared formulation lets one model cover\nstructured visual understanding, dense geometric prediction, segmentation, and\nmulti-view visual geometry while keeping outputs decodable for standard\nbenchmarks.\n\nTo enable large-scale training, we convert heterogeneous computer-vision\nannotations into instruction-response examples and construct the\n**SenseNova-Vision Corpus**, spanning decodable text, image, and mixed\ntext-image targets. Starting from an off-the-shelf pretrained UMM,\nSenseNova-Vision is trained primarily on this corpus, with auxiliary multimodal\ndata used to preserve general understanding and generation capability, and\nrequires no task-specific prediction heads, decoders, or architectural branches.\n\n### 🏗️ Key Contributions\n\n- 🔗 We introduce a unified multimodal generation formulation that casts heterogeneous computer vision tasks into the native input-output spaces of UMMs.\n- 🧩 We construct the SenseNova-Vision Corpus, a large-scale computer-vision instruction-response corpus with decodable text, image, and mixed text-image targets.\n- ✨ We train SenseNova-Vision and show strong results across structured visual understanding, dense geometric prediction, segmentation, and multi-view visual geometry, while supporting language-defined task variants beyond fixed benchmark schemas.\n\n## 🛠️ Quick Start\n\nThis repository provides one entrypoint for examples, single-image inference,\ninteractive inference, and benchmark inference. For the full runtime guide, see\n[`docs\u002FEVAL.md`](.\u002Fdocs\u002FEVAL.md).\n\nCreate the environment from the repository root:\n\n```bash\ngit clone https:\u002F\u002Fgithub.com\u002FOpenSenseNova\u002FSenseNova-Vision.git\ncd SenseNova-Vision\nbash setup.sh sensenova-vision\nconda activate sensenova-vision\n```\n\nRun the curated example:\n\n```bash\nbash scripts\u002Frun_sensenova_vision.sh example\n```\n\nRun one inference request:\n\n```bash\nbash scripts\u002Frun_sensenova_vision.sh inference \\\n  binary_seg \\\n  \"person\" \\\n  examples\u002Fimages\u002F2.jpg\n```\n\nLaunch the web demo. The wrapper prints the local URL before starting Gradio.\n**Recommended:** 1 x 80GB GPU for the full web demo.\n\n```bash\nMODEL_PATH=\u002Fpath\u002Fto\u002FSenseNova-Vision-7B-MoT \\\n  bash scripts\u002Frun_sensenova_vision.sh demo\n```\n\nRun the full benchmark after preparing `datas\u002F` and `jsonl_generate\u002F` according\nto [`docs\u002Fdata_prepare.md`](.\u002Fdocs\u002Fdata_prepare.md).\n**Recommended:** at least one 8 x 80GB GPU machine for the full benchmark.\n\n```bash\nbash scripts\u002Frun_sensenova_vision.sh benchmark all \\\n  --num_gpus 8 \\\n  --tasks_per_gpu 2\n```\n\n## 🏆 Benchmark Results\n\nSenseNova-Vision is evaluated across structured visual understanding, dense geometric prediction, segmentation, and multi-view visual geometry. All tasks are formulated with natural-language instructions: textual outputs are parsed into benchmark-specific structures such as boxes, points, recognized text, keypoints, and camera parameters, while image outputs are decoded into masks, depth maps, normal maps, or 3D point maps.\n\n### Structured Visual Understanding\n\nStructured visual understanding evaluates tasks whose outputs can be represented as structured textual predictions, such as bounding boxes, points, recognized text, and keypoint coordinates.\n\n\u003Ctable>\n  \u003Cthead>\n    \u003Ctr>\n      \u003Cth align=\"center\" rowspan=\"3\">Method\u003C\u002Fth>\n      \u003Cth align=\"center\" colspan=\"6\">Object Detection\u003C\u002Fth>\n      \u003Cth align=\"center\" colspan=\"2\">OCR\u003C\u002Fth>\n      \u003Cth align=\"center\">GUI\u003C\u002Fth>\n      \u003Cth align=\"center\">Keypoint\u003C\u002Fth>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Cth align=\"center\">COCO-Com.\u003C\u002Fth>\n      \u003Cth align=\"center\">HR\u002FRefCOCOg V\u002FT\u003C\u002Fth>\n      \u003Cth align=\"center\">LVIS\u003C\u002Fth>\n      \u003Cth align=\"center\">Dense200\u003C\u002Fth>\n      \u003Cth align=\"center\" colspan=\"2\">VisDrone\u003C\u002Fth>\n      \u003Cth align=\"center\">HierText\u003C\u002Fth>\n      \u003Cth align=\"center\">ICDAR15\u003C\u002Fth>\n      \u003Cth align=\"center\">ScreenSpot-V2\u003C\u002Fth>\n      \u003Cth align=\"center\">COCO-Kpt.\u003C\u002Fth>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Cth align=\"center\">bbox\u003C\u002Fth>\n      \u003Cth align=\"center\">bbox\u003C\u002Fth>\n      \u003Cth align=\"center\">bbox\u003C\u002Fth>\n      \u003Cth align=\"center\">bbox\u003C\u002Fth>\n      \u003Cth align=\"center\">bbox\u003C\u002Fth>\n      \u003Cth align=\"center\">point\u003C\u002Fth>\n      \u003Cth align=\"center\">bbox\u003C\u002Fth>\n      \u003Cth align=\"center\">bbox\u003C\u002Fth>\n      \u003Cth align=\"center\">bbox\u003C\u002Fth>\n      \u003Cth align=\"center\">point\u003C\u002Fth>\n    \u003C\u002Ftr>\n  \u003C\u002Fthead>\n  \u003Ctbody>\n    \u003Ctr>\n      \u003Ctd>Grounding DINO-Swin-T\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>56.6\u003C\u002Fstrong>\u003C\u002Ftd>\n      \u003Ctd>25.2 \u002F 45.9 \u002F 46.8\u003C\u002Ftd>\n      \u003Ctd>38.8\u003C\u002Ftd>\n      \u003Ctd>33.1\u003C\u002Ftd>\n      \u003Ctd>\u003Cu>38.5\u003C\u002Fu>\u003C\u002Ftd>\n      \u003Ctd>--\u003C\u002Ftd>\n      \u003Ctd>--\u003C\u002Ftd>\n      \u003Ctd>--\u003C\u002Ftd>\n      \u003Ctd>--\u003C\u002Ftd>\n      \u003Ctd>--\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>Bagel\u003C\u002Ftd>\n      \u003Ctd>50.2\u003C\u002Ftd>\n      \u003Ctd>74.6 \u002F 76.4 \u002F \u003Cu>77.8\u003C\u002Fu>\u003C\u002Ftd>\n      \u003Ctd>46.8\u003C\u002Ftd>\n      \u003Ctd>42.4\u003C\u002Ftd>\n      \u003Ctd>23.0\u003C\u002Ftd>\n      \u003Ctd>36.9\u003C\u002Ftd>\n      \u003Ctd>7.1\u003C\u002Ftd>\n      \u003Ctd>15.8\u003C\u002Ftd>\n      \u003Ctd>81.1\u003C\u002Ftd>\n      \u003Ctd>--\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>Qwen3-VL-8B-Instruct\u003C\u002Ftd>\n      \u003Ctd>46.6\u003C\u002Ftd>\n      \u003Ctd>70.4 \u002F 72.3 \u002F 72.6\u003C\u002Ftd>\n      \u003Ctd>43.2\u003C\u002Ftd>\n      \u003Ctd>13.5\u003C\u002Ftd>\n      \u003Ctd>28.7\u003C\u002Ftd>\n      \u003Ctd>35.7\u003C\u002Ftd>\n      \u003Ctd>22.4\u003C\u002Ftd>\n      \u003Ctd>25.4\u003C\u002Ftd>\n      \u003Ctd>\u003Cu>90.5\u003C\u002Fu>\u003C\u002Ftd>\n      \u003Ctd>--\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>Qwen3.5-9B\u003C\u002Ftd>\n      \u003Ctd>49.3\u003C\u002Ftd>\n      \u003Ctd>71.7 \u002F 72.1 \u002F 72.6\u003C\u002Ftd>\n      \u003Ctd>43.2\u003C\u002Ftd>\n      \u003Ctd>27.5\u003C\u002Ftd>\n      \u003Ctd>26.8\u003C\u002Ftd>\n      \u003Ctd>41.7\u003C\u002Ftd>\n      \u003Ctd>19.6\u003C\u002Ftd>\n      \u003Ctd>11.4\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>92.2\u003C\u002Fstrong>\u003C\u002Ftd>\n      \u003Ctd>--\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>LocateAnything\u003C\u002Ftd>\n      \u003Ctd>\u003Cu>54.7\u003C\u002Fu>\u003C\u002Ftd>\n      \u003Ctd>78.7 \u002F \u003Cu>76.7\u003C\u002Fu> \u002F 77.6\u003C\u002Ftd>\n      \u003Ctd>\u003Cu>50.7\u003C\u002Fu>\u003C\u002Ftd>\n      \u003Ctd>\u003Cu>58.7\u003C\u002Fu>\u003C\u002Ftd>\n      \u003Ctd>\u003Cu>39.9\u003C\u002Fu>\u003C\u002Ftd>\n      \u003Ctd>\u003Cu>60.4\u003C\u002Fu>\u003C\u002Ftd>\n      \u003Ctd>\u003Cu>29.1\u003C\u002Fu>\u003C\u002Ftd>\n      \u003Ctd>26.4\u003C\u002Ftd>\n      \u003Ctd>85.5\u003C\u002Ftd>\n      \u003Ctd>--\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>Rex-Omni\u003C\u002Ftd>\n      \u003Ctd>52.9\u003C\u002Ftd>\n      \u003Ctd>\u003Cu>79.9\u003C\u002Fu> \u002F 73.6 \u002F 74.3\u003C\u002Ftd>\n      \u003Ctd>46.9\u003C\u002Ftd>\n      \u003Ctd>58.3\u003C\u002Ftd>\n      \u003Ctd>35.8\u003C\u002Ftd>\n      \u003Ctd>58.9\u003C\u002Ftd>\n      \u003Ctd>28.0\u003C\u002Ftd>\n      \u003Ctd>\u003Cu>28.1\u003C\u002Fu>\u003C\u002Ftd>\n      \u003Ctd>88.4\u003C\u002Ftd>\n      \u003Ctd>\u003Cu>32.6\u003C\u002Fu>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>SenseNova-Vision\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>56.6\u003C\u002Fstrong>\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>80.2\u003C\u002Fstrong> \u002F \u003Cstrong>79.6\u003C\u002Fstrong> \u002F \u003Cstrong>80.5\u003C\u002Fstrong>\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>54.8\u003C\u002Fstrong>\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>66.8\u003C\u002Fstrong>\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>43.3\u003C\u002Fstrong>\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>62.9\u003C\u002Fstrong>\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>31.2\u003C\u002Fstrong>\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>49.5\u003C\u002Fstrong>\u003C\u002Ftd>\n      \u003Ctd>85.9\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>34.6\u003C\u002Fstrong>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n### Dense Geometric Prediction\n\nDense geometric prediction evaluates pixel-aligned geometric outputs, including monocular depth estimation and surface normal estimation.\n\n\u003Ctable>\n  \u003Cthead>\n    \u003Ctr>\n      \u003Cth align=\"center\" rowspan=\"3\">Method\u003C\u002Fth>\n      \u003Cth align=\"center\" colspan=\"5\">Depth\u003C\u002Fth>\n      \u003Cth align=\"center\" colspan=\"3\">Normal\u003C\u002Fth>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Cth align=\"center\">NYUv2\u003C\u002Fth>\n      \u003Cth align=\"center\">KITTI\u003C\u002Fth>\n      \u003Cth align=\"center\">ETH3D\u003C\u002Fth>\n      \u003Cth align=\"center\">ScanNet\u003C\u002Fth>\n      \u003Cth align=\"center\">DIODE\u003C\u002Fth>\n      \u003Cth align=\"center\">ScanNet\u003C\u002Fth>\n      \u003Cth align=\"center\">iBims-1\u003C\u002Fth>\n      \u003Cth align=\"center\">NYUv2\u003C\u002Fth>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Cth align=\"center\" colspan=\"5\">AbsRel↓ \u002F δ1↑\u003C\u002Fth>\n      \u003Cth align=\"center\" colspan=\"3\">Mean↓ \u002F 11.25°↑\u003C\u002Fth>\n    \u003C\u002Ftr>\n  \u003C\u002Fthead>\n  \u003Ctbody>\n    \u003Ctr>\n      \u003Ctd>DSINE\u003C\u002Ftd>\n      \u003Ctd>--\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\n      \u003Ctd>16.2 \u002F 61.0\u003C\u002Ftd>\u003Ctd>17.1 \u002F 67.4\u003C\u002Ftd>\u003Ctd>16.4 \u002F 59.6\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>DepthAnything\u003C\u002Ftd>\n      \u003Ctd>4.3 \u002F \u003Cstrong>98.1\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>7.6 \u002F 94.7\u003C\u002Ftd>\u003Ctd>12.7 \u002F 88.2\u003C\u002Ftd>\u003Ctd>4.3 \u002F 98.1\u003C\u002Ftd>\u003Ctd>26.0 \u002F 75.9\u003C\u002Ftd>\n      \u003Ctd>--\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>DepthAnything V2\u003C\u002Ftd>\n      \u003Ctd>4.5 \u002F 97.9\u003C\u002Ftd>\u003Ctd>7.4 \u002F 94.6\u003C\u002Ftd>\u003Ctd>13.1 \u002F 86.5\u003C\u002Ftd>\u003Ctd>4.2 \u002F 97.8\u003C\u002Ftd>\u003Ctd>26.5 \u002F 73.4\u003C\u002Ftd>\n      \u003Ctd>--\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>*MoGe-2\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>3.5\u003C\u002Fstrong> \u002F 98.0\u003C\u002Ftd>\u003Ctd>\u003Cstrong>5.5\u003C\u002Fstrong> \u002F \u003Cstrong>97.7\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>3.4\u003C\u002Fstrong> \u002F \u003Cstrong>98.8\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>3.4\u003C\u002Fstrong> \u002F \u003Cstrong>98.3\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>23.0\u003C\u002Fstrong> \u002F \u003Cstrong>82.3\u003C\u002Fstrong>\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>12.8\u003C\u002Fstrong> \u002F \u003Cstrong>68.4\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>14.7\u003C\u002Fstrong> \u002F \u003Cstrong>70.4\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>14.7\u003C\u002Fstrong> \u002F \u003Cstrong>62.3\u003C\u002Fstrong>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>Marigold\u003C\u002Ftd>\n      \u003Ctd>5.5 \u002F 96.4\u003C\u002Ftd>\u003Ctd>9.9 \u002F 91.6\u003C\u002Ftd>\u003Ctd>6.5 \u002F 95.9\u003C\u002Ftd>\u003Ctd>6.4 \u002F 95.2\u003C\u002Ftd>\u003Ctd>30.8 \u002F \u003Cu>77.3\u003C\u002Fu>\u003C\u002Ftd>\n      \u003Ctd>21.3 \u002F 45.6\u003C\u002Ftd>\u003Ctd>18.5 \u002F 64.7\u003C\u002Ftd>\u003Ctd>20.9 \u002F 50.5\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>DICEPTION\u003C\u002Ftd>\n      \u003Ctd>6.1 \u002F 96.0\u003C\u002Ftd>\u003Ctd>6.9 \u002F 94.9\u003C\u002Ftd>\u003Ctd>5.0 \u002F 97.5\u003C\u002Ftd>\u003Ctd>7.2 \u002F 94.4\u003C\u002Ftd>\u003Ctd>28.9 \u002F 72.2\u003C\u002Ftd>\n      \u003Ctd>18.8 \u002F 53.6\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>18.3 \u002F 52.9\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>FE2E\u003C\u002Ftd>\n      \u003Ctd>\u003Cu>4.1\u003C\u002Fu> \u002F \u003Cu>97.7\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd>\u003Cu>6.6\u003C\u002Fu> \u002F \u003Cstrong>96.0\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>3.8\u003C\u002Fstrong> \u002F \u003Cstrong>98.7\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>4.4 \u002F 97.5\u003C\u002Ftd>\u003Ctd>22.8 \u002F \u003Cstrong>81.2\u003C\u002Fstrong>\u003C\u002Ftd>\n      \u003Ctd>\u003Cu>13.8\u003C\u002Fu> \u002F \u003Cu>67.2\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>15.1\u003C\u002Fstrong> \u002F \u003Cstrong>70.6\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cu>16.2\u003C\u002Fu> \u002F \u003Cu>59.6\u003C\u002Fu>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>Lotus-2\u003C\u002Ftd>\n      \u003Ctd>\u003Cu>4.1\u003C\u002Fu> \u002F 97.6\u003C\u002Ftd>\u003Ctd>6.7 \u002F 94.5\u003C\u002Ftd>\u003Ctd>4.6 \u002F \u003Cu>98.1\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd>\u003Cu>4.2\u003C\u002Fu> \u002F \u003Cu>97.6\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd>\u003Cu>22.1\u003C\u002Fu> \u002F 75.2\u003C\u002Ftd>\n      \u003Ctd>14.2 \u002F 66.8\u003C\u002Ftd>\u003Ctd>\u003Cu>15.4\u003C\u002Fu> \u002F \u003Cu>70.4\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd>16.9 \u002F 59.0\u003C\u002Ftd>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Ctd>SenseNova-Vision\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>4.0\u003C\u002Fstrong> \u002F \u003Cstrong>98.1\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>5.9\u003C\u002Fstrong> \u002F \u003Cu>95.9\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd>\u003Cu>4.3\u003C\u002Fu> \u002F 97.4\u003C\u002Ftd>\u003Ctd>\u003Cstrong>3.9\u003C\u002Fstrong> \u002F \u003Cstrong>98.0\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>20.6\u003C\u002Fstrong> \u002F 76.4\u003C\u002Ftd>\n      \u003Ctd>\u003Cstrong>12.8\u003C\u002Fstrong> \u002F \u003Cstrong>68.9\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cu>15.4\u003C\u002Fu> \u002F 69.1\u003C\u002Ftd>\u003Ctd>\u003Cstrong>14.4\u003C\u002Fstrong> \u002F \u003Cstrong>62.7\u003C\u002Fstrong>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n### Segmentation\n\nSegmentation evaluates mask prediction under semantic, referring, reasoning, grounded, and interactive guidance.\n\n\u003Ctable>\n  \u003Cthead>\n    \u003Ctr>\n      \u003Cth align=\"center\" rowspan=\"2\">Method\u003C\u002Fth>\n      \u003Cth align=\"center\">Gen. Seg.\u003C\u002Fth>\n      \u003Cth align=\"center\">Ref. Seg.\u003C\u002Fth>\n      \u003Cth align=\"center\">Rea. Seg.\u003C\u002Fth>\n      \u003Cth align=\"center\">GCG Seg.\u003C\u002Fth>\n      \u003Cth align=\"center\">Inter. Seg.\u003C\u002Fth>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Cth align=\"center\">Pan. \u002F Sem.\u003C\u002Fth>\n      \u003Cth align=\"center\">RefCOCO \u002F + \u002F g\u003C\u002Fth>\n      \u003Cth align=\"center\">Val \u002F Test\u003C\u002Fth>\n      \u003Cth align=\"center\">Val \u002F Test\u003C\u002Fth>\n      \u003Cth align=\"center\">Point \u002F Box\u003C\u002Fth>\n    \u003C\u002Ftr>\n  \u003C\u002Fthead>\n  \u003Ctbody>\n    \u003Ctr>\u003Ctd>LISA-7B\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>74.9 \u002F 65.1 \u002F 67.9\u003C\u002Ftd>\u003Ctd>52.9 \u002F 47.3\u003C\u002Ftd>\u003Ctd>62.0 \u002F 61.7\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003C\u002Ftr>\n    \u003Ctr>\u003Ctd>PSALM\u003C\u002Ftd>\u003Ctd>\u003Cstrong>55.9\u003C\u002Fstrong> \u002F \u003Cstrong>66.6\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>83.6 \u002F 72.9 \u002F 73.8\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>\u003Cu>64.3\u003C\u002Fu> \u002F 67.3\u003C\u002Ftd>\u003C\u002Ftr>\n    \u003Ctr>\u003Ctd>Text4Seg\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>79.2 \u002F 72.8 \u002F 74.0\u003C\u002Ftd>\u003Ctd>59.1 \u002F 57.1\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003C\u002Ftr>\n    \u003Ctr>\u003Ctd>LENS\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>\u003Cu>84.2\u003C\u002Fu> \u002F \u003Cstrong>79.4\u003C\u002Fstrong> \u002F \u003Cu>81.2\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd>\u003Cu>62.1\u003C\u002Fu> \u002F 57.2\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003C\u002Ftr>\n    \u003Ctr>\u003Ctd>ConverSeg\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>79.4 \u002F 74.3 \u002F 74.9\u003C\u002Ftd>\u003Ctd>61.9 \u002F 57.0\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003Ctd>--\u003C\u002Ftd>\u003C\u002Ftr>\n    \u003Ctr>\u003Ctd>X-SAM\u003C\u002Ftd>\u003Ctd>\u003Cu>54.7\u003C\u002Fu> \u002F \u003Cu>66.5\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>85.1\u003C\u002Fstrong> \u002F \u003Cu>78.0\u003C\u002Fu> \u002F \u003Cstrong>83.8\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>56.6 \u002F \u003Cu>57.8\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>69.4\u003C\u002Fstrong> \u002F \u003Cstrong>69.0\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>65.4\u003C\u002Fstrong> \u002F \u003Cu>70.0\u003C\u002Fu>\u003C\u002Ftd>\u003C\u002Ftr>\n    \u003Ctr>\u003Ctd>SenseNova-Vision\u003C\u002Ftd>\u003Ctd>48.8 \u002F 64.0\u003C\u002Ftd>\u003Ctd>81.3 \u002F 76.0 \u002F 80.3\u003C\u002Ftd>\u003Ctd>\u003Cstrong>63.2\u003C\u002Fstrong> \u002F \u003Cstrong>60.7\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cu>65.7\u003C\u002Fu> \u002F \u003Cu>66.2\u003C\u002Fu>\u003C\u002Ftd>\u003Ctd>60.9 \u002F \u003Cstrong>73.9\u003C\u002Fstrong>\u003C\u002Ftd>\u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n### Multi-View Visual Geometry\n\nMulti-view visual geometry evaluates geometric prediction from multiple input images, including multi-view point map reconstruction and camera pose estimation.\n\n\u003Ctable>\n  \u003Cthead>\n    \u003Ctr>\n      \u003Cth align=\"center\" rowspan=\"3\">Method\u003C\u002Fth>\n      \u003Cth align=\"center\" colspan=\"2\">Multi-View Reconstruction\u003C\u002Fth>\n      \u003Cth align=\"center\" colspan=\"2\">Camera Pose\u003C\u002Fth>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Cth align=\"center\" colspan=\"2\">Acc.↓ \u002F Comp.↓ \u002F F1↑\u003C\u002Fth>\n      \u003Cth align=\"center\" colspan=\"2\">RRA@30↑ \u002F RTA@30↑ \u002F AUC@30↑\u003C\u002Fth>\n    \u003C\u002Ftr>\n    \u003Ctr>\n      \u003Cth align=\"center\">7Scenes\u003C\u002Fth>\n      \u003Cth align=\"center\">ETH3D\u003C\u002Fth>\n      \u003Cth align=\"center\">Re10K\u003C\u002Fth>\n      \u003Cth align=\"center\">CO3Dv2\u003C\u002Fth>\n    \u003C\u002Ftr>\n  \u003C\u002Fthead>\n  \u003Ctbody>\n    \u003Ctr>\u003Ctd>DUSt3R\u003C\u002Ftd>\u003Ctd>0.026 \u002F 0.034 \u002F 87.1\u003C\u002Ftd>\u003Ctd>0.359 \u002F 0.531 \u002F 66.6\u003C\u002Ftd>\u003Ctd>99.8 \u002F 84.9 \u002F 67.6\u003C\u002Ftd>\u003Ctd>97.7 \u002F 93.4 \u002F 78.3\u003C\u002Ftd>\u003C\u002Ftr>\n    \u003Ctr>\u003Ctd>DepthAnything3\u003C\u002Ftd>\u003Ctd>\u003Cstrong>0.020\u003C\u002Fstrong> \u002F \u003Cstrong>0.026\u003C\u002Fstrong> \u002F \u003Cstrong>90.5\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>0.228 \u002F 0.212 \u002F 76.6\u003C\u002Ftd>\u003Ctd>\u003Cstrong>100.0\u003C\u002Fstrong> \u002F \u003Cstrong>96.4\u003C\u002Fstrong> \u002F \u003Cstrong>89.6\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>99.3\u003C\u002Fstrong> \u002F \u003Cstrong>98.0\u003C\u002Fstrong> \u002F \u003Cstrong>91.8\u003C\u002Fstrong>\u003C\u002Ftd>\u003C\u002Ftr>\n    \u003Ctr>\u003Ctd>VGGT\u003C\u002Ftd>\u003Ctd>0.023 \u002F 0.032 \u002F 88.4\u003C\u002Ftd>\u003Ctd>\u003Cstrong>0.177\u003C\u002Fstrong> \u002F \u003Cstrong>0.155\u003C\u002Fstrong> \u002F \u003Cstrong>80.9\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>100.0\u003C\u002Fstrong> \u002F 93.5 \u002F 79.3\u003C\u002Ftd>\u003Ctd>98.3 \u002F 96.6 \u002F 89.2\u003C\u002Ftd>\u003C\u002Ftr>\n    \u003Ctr>\u003Ctd>MoRe\u003C\u002Ftd>\u003Ctd>0.038 \u002F 0.039 \u002F 77.1\u003C\u002Ftd>\u003Ctd>0.348 \u002F 0.318 \u002F 62.7\u003C\u002Ftd>\u003Ctd>\u003Cstrong>100.0\u003C\u002Fstrong> \u002F 94.0 \u002F 79.1\u003C\u002Ftd>\u003Ctd>98.4 \u002F 96.3 \u002F 83.0\u003C\u002Ftd>\u003C\u002Ftr>\n    \u003Ctr>\u003Ctd>MapAnything\u003C\u002Ftd>\u003Ctd>\u003Cstrong>0.027\u003C\u002Fstrong> \u002F 0.029 \u002F 87.8\u003C\u002Ftd>\u003Ctd>0.400 \u002F 0.524 \u002F 67.0\u003C\u002Ftd>\u003Ctd>\u003Cstrong>100.0\u003C\u002Fstrong> \u002F 93.5 \u002F \u003Cstrong>80.7\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>95.5 \u002F 91.6 \u002F 70.9\u003C\u002Ftd>\u003C\u002Ftr>\n    \u003Ctr>\u003Ctd>G2VLM\u003C\u002Ftd>\u003Ctd>0.084 \u002F 0.056 \u002F 59.2\u003C\u002Ftd>\u003Ctd>0.784 \u002F 0.553 \u002F 36.7\u003C\u002Ftd>\u003Ctd>99.8 \u002F 77.5 \u002F 51.8\u003C\u002Ftd>\u003Ctd>96.3 \u002F 92.0 \u002F 55.2\u003C\u002Ftd>\u003C\u002Ftr>\n    \u003Ctr>\u003Ctd>SenseNova-Vision\u003C\u002Ftd>\u003Ctd>0.028 \u002F \u003Cstrong>0.026\u003C\u002Fstrong> \u002F \u003Cstrong>87.9\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>0.301\u003C\u002Fstrong> \u002F \u003Cstrong>0.175\u003C\u002Fstrong> \u002F \u003Cstrong>72.2\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>99.8 \u002F \u003Cstrong>94.2\u003C\u002Fstrong> \u002F 77.3\u003C\u002Ftd>\u003Ctd>\u003Cstrong>97.4\u003C\u002Fstrong> \u002F \u003Cstrong>95.4\u003C\u002Fstrong> \u002F \u003Cstrong>80.1\u003C\u002Fstrong>\u003C\u002Ftd>\u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n### Comparison with Generalist Vision Models\n\nWe further compare SenseNova-Vision with recent generalist visual models that span multiple visual capabilities.\n\n\u003Ctable>\n  \u003Cthead>\n    \u003Ctr>\u003Cth align=\"center\" rowspan=\"3\">Method\u003C\u002Fth>\u003Cth align=\"center\">Detection\u003C\u002Fth>\u003Cth align=\"center\">Sem. Seg.\u003C\u002Fth>\u003Cth align=\"center\">Ref. Seg.\u003C\u002Fth>\u003Cth align=\"center\">Depth\u003C\u002Fth>\u003C\u002Ftr>\n    \u003Ctr>\u003Cth align=\"center\">mAP\u003C\u002Fth>\u003Cth align=\"center\">mIoU\u003C\u002Fth>\u003Cth align=\"center\">cIoU\u003C\u002Fth>\u003Cth align=\"center\">δ1\u003C\u002Fth>\u003C\u002Ftr>\n    \u003Ctr>\u003Cth align=\"center\">COCO\u003C\u002Fth>\u003Cth align=\"center\">Cityscapes\u003C\u002Fth>\u003Cth align=\"center\">RefCOCO \u002F + \u002F g\u003C\u002Fth>\u003Cth align=\"center\">NYUv2\u003C\u002Fth>\u003C\u002Ftr>\n  \u003C\u002Fthead>\n  \u003Ctbody>\n    \u003Ctr>\u003Ctd>Youtu-VL\u003C\u002Ftd>\u003Ctd>47.1\u003C\u002Ftd>\u003Ctd>70.4\u003C\u002Ftd>\u003Ctd>80.7 \u002F \u003Cstrong>76.2\u003C\u002Fstrong> \u002F 76.5\u003C\u002Ftd>\u003Ctd>90.4\u003C\u002Ftd>\u003C\u002Ftr>\n    \u003Ctr>\u003Ctd>SenseNova-Vision\u003C\u002Ftd>\u003Ctd>\u003Cstrong>53.7\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>71.2\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>81.3\u003C\u002Fstrong> \u002F 76.0 \u002F \u003Cstrong>80.3\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>98.1\u003C\u002Fstrong>\u003C\u002Ftd>\u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n\u003Ctable>\n  \u003Cthead>\n    \u003Ctr>\u003Cth align=\"center\" rowspan=\"3\">Method\u003C\u002Fth>\u003Cth align=\"center\">Sem. Seg.\u003C\u002Fth>\u003Cth align=\"center\">Ref. Seg.\u003C\u002Fth>\u003Cth align=\"center\">Rea. Seg.\u003C\u002Fth>\u003Cth align=\"center\" colspan=\"4\">Depth\u003C\u002Fth>\u003Cth align=\"center\" colspan=\"3\">Normal\u003C\u002Fth>\u003C\u002Ftr>\n    \u003Ctr>\u003Cth align=\"center\">mIoU\u003C\u002Fth>\u003Cth align=\"center\">cIoU\u003C\u002Fth>\u003Cth align=\"center\">gIoU\u003C\u002Fth>\u003Cth align=\"center\" colspan=\"4\">δ1\u003C\u002Fth>\u003Cth align=\"center\" colspan=\"3\">Mean Error↓\u003C\u002Fth>\u003C\u002Ftr>\n    \u003Ctr>\u003Cth align=\"center\">Cityscapes\u003C\u002Fth>\u003Cth align=\"center\">RefCOCOg\u003C\u002Fth>\u003Cth align=\"center\">ReasonSeg\u003C\u002Fth>\u003Cth align=\"center\">KITTI\u003C\u002Fth>\u003Cth align=\"center\">NYUv2\u003C\u002Fth>\u003Cth align=\"center\">DIODE\u003C\u002Fth>\u003Cth align=\"center\">ETH3D\u003C\u002Fth>\u003Cth align=\"center\">NYUv2\u003C\u002Fth>\u003Cth align=\"center\">ScanNet\u003C\u002Fth>\u003Cth align=\"center\">DIODE\u003C\u002Fth>\u003C\u002Ftr>\n  \u003C\u002Fthead>\n  \u003Ctbody>\n    \u003Ctr>\u003Ctd>Vision Banana\u003C\u002Ftd>\u003Ctd>69.9\u003C\u002Ftd>\u003Ctd>73.8\u003C\u002Ftd>\u003Ctd>79.3\u003C\u002Ftd>\u003Ctd>91.5\u003C\u002Ftd>\u003Ctd>94.8\u003C\u002Ftd>\u003Ctd>91.7\u003C\u002Ftd>\u003Ctd>93.5\u003C\u002Ftd>\u003Ctd>17.8\u003C\u002Ftd>\u003Ctd>15.1\u003C\u002Ftd>\u003Ctd>\u003Cstrong>13.8\u003C\u002Fstrong>\u003C\u002Ftd>\u003C\u002Ftr>\n    \u003Ctr>\u003Ctd>SenseNova-Vision\u003C\u002Ftd>\u003Ctd>\u003Cstrong>71.2\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>80.3\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>63.2\u003C\u002Ftd>\u003Ctd>95.9\u003C\u002Ftd>\u003Ctd>98.1\u003C\u002Ftd>\u003Ctd>76.4\u003C\u002Ftd>\u003Ctd>97.4\u003C\u002Ftd>\u003Ctd>\u003Cstrong>14.4\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>\u003Cstrong>12.8\u003C\u002Fstrong>\u003C\u002Ftd>\u003Ctd>15.3\u003C\u002Ftd>\u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n## 🎨 Showcase\n\n\u003Cp align=\"center\">\n  \u003Ca href=\".\u002Fassets\u002Fshowcase\u002Ffig7_sensenova_vision_results.webp\">\u003Cimg src=\".\u002Fassets\u002Fshowcase\u002Ffig7_sensenova_vision_results.webp\" alt=\"SenseNova-Vision qualitative results across vision tasks\" width=\"900\">\u003C\u002Fa>\n\u003C\u002Fp>\n\n\u003Cdetails>\n\u003Csummary>Object Detection\u003C\u002Fsummary>\n\n\u003Ctable align=\"center\">\n  \u003Cthead>\n    \u003Ctr>\n      \u003Cth align=\"center\">COCO-Com.\u003C\u002Fth>\n      \u003Cth align=\"center\">LVIS\u003C\u002Fth>\n      \u003Cth align=\"center\">Dense200\u003C\u002Fth>\n      \u003Cth align=\"center\">VisDrone\u003C\u002Fth>\n    \u003C\u002Ftr>\n  \u003C\u002Fthead>\n  \u003Ctbody>\n    \u003Ctr>\n      \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fobject_detection\u002Fcoco_common.jpg\">\u003Cimg height=\"180\" alt=\"common object detection COCO case\" src=\".\u002Fassets\u002Fshowcase\u002Fobject_detection\u002Fcoco_common.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n      \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fobject_detection\u002FLIVS.jpg\">\u003Cimg height=\"180\" alt=\"long-tail object detection LVIS case\" src=\".\u002Fassets\u002Fshowcase\u002Fobject_detection\u002FLIVS.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n      \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fobject_detection\u002Fdense200.jpg\">\u003Cimg height=\"180\" alt=\"dense object detection Dense200 case\" src=\".\u002Fassets\u002Fshowcase\u002Fobject_detection\u002Fdense200.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n      \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fobject_detection\u002Fvisdrone.jpg\">\u003Cimg height=\"180\" alt=\"object detection VisDrone case\" src=\".\u002Fassets\u002Fshowcase\u002Fobject_detection\u002Fvisdrone.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003C\u002Ftr>\n  \u003C\u002Ftbody>\n\u003C\u002Ftable>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Referring Detection\u003C\u002Fsummary>\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Freferring_detection\u002F141964854306.jpg\">\u003Cimg height=\"260\" alt=\"referring detection case 1\" src=\".\u002Fassets\u002Fshowcase\u002Freferring_detection\u002F141964854306.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Freferring_detection\u002FCOCO_train2014_000000204294.jpg\">\u003Cimg height=\"260\" alt=\"referring detection case 2\" src=\".\u002Fassets\u002Fshowcase\u002Freferring_detection\u002FCOCO_train2014_000000204294.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>OCR\u003C\u002Fsummary>\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Cth align=\"center\" colspan=\"2\">Textline Level\u003C\u002Fth>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Focr\u002Ftextline_1.jpg\">\u003Cimg height=\"220\" alt=\"OCR textline case 1\" src=\".\u002Fassets\u002Fshowcase\u002Focr\u002Ftextline_1.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Focr\u002Ftextline_2.jpg\">\u003Cimg height=\"220\" alt=\"OCR textline case 2\" src=\".\u002Fassets\u002Fshowcase\u002Focr\u002Ftextline_2.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Cth align=\"center\" colspan=\"2\">Word Level\u003C\u002Fth>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Focr\u002Fword_1.jpg\">\u003Cimg height=\"220\" alt=\"OCR word case 1\" src=\".\u002Fassets\u002Fshowcase\u002Focr\u002Fword_1.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Focr\u002Fword_2.jpg\">\u003Cimg height=\"220\" alt=\"OCR word case 2\" src=\".\u002Fassets\u002Fshowcase\u002Focr\u002Fword_2.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Visual Prompting\u003C\u002Fsummary>\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fvisual_prompt\u002F2009.jpg\">\u003Cimg height=\"220\" alt=\"visual prompt bbox case 1\" src=\".\u002Fassets\u002Fshowcase\u002Fvisual_prompt\u002F2009.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fvisual_prompt\u002F2271.jpg\">\u003Cimg height=\"220\" alt=\"visual prompt bbox case 2\" src=\".\u002Fassets\u002Fshowcase\u002Fvisual_prompt\u002F2271.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fvisual_prompt\u002F2916.jpg\">\u003Cimg height=\"220\" alt=\"visual prompt bbox case 3\" src=\".\u002Fassets\u002Fshowcase\u002Fvisual_prompt\u002F2916.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fvisual_prompt\u002F7082.jpg\">\u003Cimg height=\"220\" alt=\"visual prompt bbox case 4\" src=\".\u002Fassets\u002Fshowcase\u002Fvisual_prompt\u002F7082.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Layout Grounding\u003C\u002Fsummary>\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Flayout_grounding\u002Flayout_1.jpg\">\u003Cimg height=\"240\" alt=\"layout grounding case 1\" src=\".\u002Fassets\u002Fshowcase\u002Flayout_grounding\u002Flayout_1.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Flayout_grounding\u002Flayout_2.jpg\">\u003Cimg height=\"240\" alt=\"layout grounding case 2\" src=\".\u002Fassets\u002Fshowcase\u002Flayout_grounding\u002Flayout_2.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Keypoint Detection\u003C\u002Fsummary>\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Cth align=\"center\" colspan=\"2\">Human\u003C\u002Fth>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fkeypoints\u002Fhuman_1.jpg\">\u003Cimg height=\"240\" alt=\"human keypoint case 1\" src=\".\u002Fassets\u002Fshowcase\u002Fkeypoints\u002Fhuman_1.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fkeypoints\u002Fhuman_2.jpg\">\u003Cimg height=\"240\" alt=\"human keypoint case 2\" src=\".\u002Fassets\u002Fshowcase\u002Fkeypoints\u002Fhuman_2.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Cth align=\"center\" colspan=\"2\">Animal\u003C\u002Fth>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fkeypoints\u002Fanimal_1.jpg\">\u003Cimg height=\"240\" alt=\"animal keypoint case 1\" src=\".\u002Fassets\u002Fshowcase\u002Fkeypoints\u002Fanimal_1.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fkeypoints\u002Fanimal_2.jpg\">\u003Cimg height=\"240\" alt=\"animal keypoint case 2\" src=\".\u002Fassets\u002Fshowcase\u002Fkeypoints\u002Fanimal_2.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>GUI Grounding\u003C\u002Fsummary>\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fgui_grounding\u002Fgui_1.jpg\">\u003Cimg height=\"220\" alt=\"GUI grounding case 1\" src=\".\u002Fassets\u002Fshowcase\u002Fgui_grounding\u002Fgui_1.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fgui_grounding\u002Fgui_2.jpg\">\u003Cimg height=\"220\" alt=\"GUI grounding case 2\" src=\".\u002Fassets\u002Fshowcase\u002Fgui_grounding\u002Fgui_2.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Dense Geometric Prediction\u003C\u002Fsummary>\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fdense_geometry\u002Fdense_geometry.jpg\">\u003Cimg width=\"900\" alt=\"dense geometric prediction depth and normal cases\" src=\".\u002Fassets\u002Fshowcase\u002Fdense_geometry\u002Fdense_geometry.jpg\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003C\u002Fdetails>\n\u003Cdetails>\n\u003Csummary>Panoptic Segmentation\u003C\u002Fsummary>\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fpan\u002F000000439525_panoptic.webp\">\u003Cimg height=\"240\" alt=\"panoptic segmentation case 1\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fpan\u002F000000439525_panoptic.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fpan\u002F000000563603_panoptic.webp\">\u003Cimg height=\"240\" alt=\"panoptic segmentation case 2\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fpan\u002F000000563603_panoptic.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fpan\u002F000000237928_panoptic.webp\">\u003Cimg height=\"240\" alt=\"panoptic segmentation case 3\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fpan\u002F000000237928_panoptic.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fpan\u002F000000009772_panoptic.webp\">\u003Cimg height=\"240\" alt=\"panoptic segmentation case 4\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fpan\u002F000000009772_panoptic.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Semantic Segmentation\u003C\u002Fsummary>\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fsem\u002F000000001000_semantic.webp\">\u003Cimg height=\"240\" alt=\"semantic segmentation case 1\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fsem\u002F000000001000_semantic.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fsem\u002F000000017627_semantic.webp\">\u003Cimg height=\"240\" alt=\"semantic segmentation case 2\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fsem\u002F000000017627_semantic.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fsem\u002F000000028993_semantic.webp\">\u003Cimg height=\"240\" alt=\"semantic segmentation case 3\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fsem\u002F000000028993_semantic.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fsem\u002F000000074733_semantic.webp\">\u003Cimg height=\"240\" alt=\"semantic segmentation case 4\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fsem\u002F000000074733_semantic.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Referring Segmentation\u003C\u002Fsummary>\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fref\u002Fsample_003607_COCO_train2014_000000084712_42623_main_center_person_pred_vis.webp\">\u003Cimg height=\"240\" alt=\"referring segmentation case 1\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fref\u002Fsample_003607_COCO_train2014_000000084712_42623_main_center_person_pred_vis.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fref\u002Fsample_003854_COCO_train2014_000000232371_30263_left_giraffe_pred_vis.webp\">\u003Cimg height=\"240\" alt=\"referring segmentation case 2\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fref\u002Fsample_003854_COCO_train2014_000000232371_30263_left_giraffe_pred_vis.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fref\u002Fsample_004283_COCO_train2014_000000388421_16726_older_man_pred_vis.webp\">\u003Cimg height=\"240\" alt=\"referring segmentation case 3\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fref\u002Fsample_004283_COCO_train2014_000000388421_16726_older_man_pred_vis.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fref\u002Fsample_010222_COCO_train2014_000000180179_34673_middle_zebra_pred_vis.webp\">\u003Cimg height=\"240\" alt=\"referring segmentation case 4\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Fref\u002Fsample_010222_COCO_train2014_000000180179_34673_middle_zebra_pred_vis.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Reasoning Segmentation\u003C\u002Fsummary>\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Frea\u002Fsample_000128_14013318558_59e559a0a5_o_in_a_music_class_students_usually_learn_to_play_various_instruments.__b2f4083db9_pred_vis.webp\">\u003Cimg height=\"240\" alt=\"reasoning segmentation case 1\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Frea\u002Fsample_000128_14013318558_59e559a0a5_o_in_a_music_class_students_usually_learn_to_play_various_instruments.__b2f4083db9_pred_vis.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Frea\u002Fsample_001270_4780863298_1e6c37d2b8_o_in_colder_seasons_when_the_weather_can_be_unpredictable_please_identi_8a93ad4d2c_pred_vis.webp\">\u003Cimg height=\"240\" alt=\"reasoning segmentation case 2\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Frea\u002Fsample_001270_4780863298_1e6c37d2b8_o_in_colder_seasons_when_the_weather_can_be_unpredictable_please_identi_8a93ad4d2c_pred_vis.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Frea\u002Fsample_001431_5183659728_546436cdcb_o_what_object_in_the_picture_could_be_utilized_as_an_accessory_worn_aro_58d636e391_pred_vis.webp\">\u003Cimg height=\"240\" alt=\"reasoning segmentation case 3\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Frea\u002Fsample_001431_5183659728_546436cdcb_o_what_object_in_the_picture_could_be_utilized_as_an_accessory_worn_aro_58d636e391_pred_vis.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Frea\u002Fsample_001998_7302072422_9c406bf68a_o_when_sailing_on_water_adjusting_the_sails_is_necessary_for_controllin_cd00167b1f_pred_vis.webp\">\u003Cimg height=\"240\" alt=\"reasoning segmentation case 4\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Frea\u002Fsample_001998_7302072422_9c406bf68a_o_when_sailing_on_water_adjusting_the_sails_is_necessary_for_controllin_cd00167b1f_pred_vis.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Grounded Conversation Generation Segmentation\u003C\u002Fsummary>\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002FGCG\u002F20260630-192539.webp\">\u003Cimg height=\"240\" alt=\"grounded conversation generation segmentation case 1\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002FGCG\u002F20260630-192539.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002FGCG\u002F20260630-201751.webp\">\u003Cimg height=\"240\" alt=\"grounded conversation generation segmentation case 2\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002FGCG\u002F20260630-201751.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002FGCG\u002F20260630-201826.webp\">\u003Cimg height=\"240\" alt=\"grounded conversation generation segmentation case 3\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002FGCG\u002F20260630-201826.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002FGCG\u002F20260630-202213.webp\">\u003Cimg height=\"240\" alt=\"grounded conversation generation segmentation case 4\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002FGCG\u002F20260630-202213.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003C\u002Fdetails>\n\n\u003Cdetails>\n\u003Csummary>Interactive Segmentation\u003C\u002Fsummary>\n\n\u003Ctable align=\"center\">\n  \u003Ctr>\n    \u003Cth align=\"center\" colspan=\"2\">Point Prompt\u003C\u002Fth>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Finter\u002Fpoint_prompt_1.webp\">\u003Cimg height=\"220\" alt=\"interactive segmentation point prompt case 1\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Finter\u002Fpoint_prompt_1.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Finter\u002F04_sample_000151_000000491497_point.webp\">\u003Cimg height=\"220\" alt=\"interactive segmentation point prompt case 2\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Finter\u002F04_sample_000151_000000491497_point.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Cth align=\"center\">Box Prompt\u003C\u002Fth>\n    \u003Cth align=\"center\">Scribble Prompt\u003C\u002Fth>\n  \u003C\u002Ftr>\n  \u003Ctr>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Finter\u002F05_sample_000071_000000480985_point.webp\">\u003Cimg height=\"220\" alt=\"interactive segmentation box prompt case 1\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Finter\u002F05_sample_000071_000000480985_point.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n    \u003Ctd align=\"center\">\u003Ca href=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Finter\u002F20260630-145454.webp\">\u003Cimg height=\"220\" alt=\"interactive segmentation scribble prompt case 1\" src=\".\u002Fassets\u002Fshowcase\u002Fsegmentation\u002Finter\u002F20260630-145454.webp\">\u003C\u002Fa>\u003C\u002Ftd>\n  \u003C\u002Ftr>\n\u003C\u002Ftable>\n\n\u003C\u002Fdetails>\n\n\n\n## Data Protocol\n\nSenseNova-Vision converts heterogeneous computer vision annotations into a common instruction-response schema. Each sample contains one or more visual inputs, a natural-language instruction that defines the task and output convention, and a decodable target represented as text, an image, or a mixed text-image response.\n\n\u003Cp align=\"center\">\n  \u003Ca href=\".\u002Fassets\u002Ffig4_training_data.webp\">\u003Cimg src=\".\u002Fassets\u002Ffig4_training_data.webp\" alt=\"Representative SenseNova-Vision data protocol examples\" width=\"900\">\u003C\u002Fa>\n\u003C\u002Fp>\n\n## ✒️ Citation\n\nIf you find SenseNova-Vision useful, please cite our technical report:\n\n```bibtex\n@misc{sensenova2026sensenovavision,\n      title={Vision as Unified Multimodal Generation}, \n      author={Xiaoyang Han and Jianhua Li and Kewang Deng and Zukai Chen and Xuanke Shi and Sihan Wang and Boxuan Li and Linyan Wang and Siyi Xie and Xin You and Jinsheng Quan and Zhongang Cai and Haiwen Diao and Ziwei Liu and Lei Yang and Dahua Lin and Quan Wang},\n      year={2026},\n      eprint={2607.06560},\n      archivePrefix={arXiv},\n      primaryClass={cs.CV},\n      url={https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.06560}, \n}\n```\n\n## License\n\nThis project is released under the [Apache 2.0 License](.\u002FLICENSE).\n","SenseNova-Vision 是一个将计算机视觉任务统一建模为多模态生成问题的开源模型框架。其核心采用统一多模态模型（UMM）架构，支持通过自然语言指令与可选视觉提示，灵活驱动文本输出（如类别、边界框、OCR、关键点）、图像输出（如分割掩码、深度图、法向量）或图文混合输出，实现检测、分割、深度估计、3D重建等多样化视觉任务的联合建模。技术上强调任务无关的生成式范式、跨任务共享参数与端到端可微训练。适用于需要轻量级统一视觉模型的研究原型开发、多任务联合推理及教育演示场景。","2026-07-10 02:30:15","CREATED_QUERY"]