Papers › Referring Image Matting

Referring Image Matting

10 Jun 2022CVPR 2023 1arXiv:2206.05149archive 2025-07-28

Jizhizi Li, Jing Zhang, DaCheng Tao

Different from conventional image matting, which either requires user-defined scribbles/trimap to extract a specific foreground object or directly extracts all the foreground objects in the image indiscriminately, we introduce a new task named Referring Image Matting (RIM) in this paper, which aims to extract the meticulous alpha matte of the specific object that best matches the given natural language description, thus enabling a more natural and simpler instruction for image matting. First, we establish a large-scale challenging dataset RefMatte by designing a comprehensive image composition and expression generation engine to automatically produce high-quality images along with diverse text attributes based on public datasets. RefMatte consists of 230 object categories, 47,500 images, 118,749 expression-region entities, and 474,996 expressions. Additionally, we construct a real-world test set with 100 high-resolution natural images and manually annotate complex phrases to evaluate the out-of-domain generalization abilities of RIM methods. Furthermore, we present a novel baseline method CLIPMat for RIM, including a context-embedded prompt, a text-driven semantic pop-up, and a multi-level details extractor. Extensive experiments on RefMatte in both keyword and expression settings validate the superiority of CLIPMat over representative methods. We hope this work could provide novel insights into image matting and encourage more follow-up studies. The dataset, code and models are available at https://github.com/JizhiziLi/RIM.

PaperPDFConference PDFCode

Code

jizhizili/rim officialmentioned in papermentioned on GitHub report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Domain GeneralizationImage MattingObjectReferring Image MattingReferring Image Matting (Expression-based)Referring Image Matting (Keyword-based)Referring Image Matting (RefMatte-RW100)Visual Grounding

Datasets

Introduced by this paper, per the archive.

RefMatte

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Referring Image Matting (Expression-based) RefMatte CLIPMat (ViT-L/14) MAD 0.0238 #1 of 4 Archive leaderboard report
Referring Image Matting (Expression-based) RefMatte CLIPMat (ViT-L/14) MAD(E) 0.0254 #1 of 4 Archive leaderboard report
Referring Image Matting (Expression-based) RefMatte CLIPMat (ViT-L/14) MSE 0.0212 #1 of 4 Archive leaderboard report
Referring Image Matting (Expression-based) RefMatte CLIPMat (ViT-L/14) MSE(E) 0.0226 #1 of 4 Archive leaderboard report
Referring Image Matting (Expression-based) RefMatte CLIPMat (ViT-L/14) SAD 42.05 #1 of 4 Archive leaderboard report
Referring Image Matting (Expression-based) RefMatte CLIPMat (ViT-L/14) SAD(E) 44.77 #1 of 4 Archive leaderboard report
Referring Image Matting (Expression-based) RefMatte CLIPMat (ViT-B/16) MAD 0.0273 #2 of 4 Archive leaderboard report
Referring Image Matting (Expression-based) RefMatte CLIPMat (ViT-B/16) MAD(E) 0.0273 #2 of 4 Archive leaderboard report
Referring Image Matting (Expression-based) RefMatte CLIPMat (ViT-B/16) MSE 0.0245 #2 of 4 Archive leaderboard report
Referring Image Matting (Expression-based) RefMatte CLIPMat (ViT-B/16) MSE(E) 0.0260 #2 of 4 Archive leaderboard report
Referring Image Matting (Expression-based) RefMatte CLIPMat (ViT-B/16) SAD 47.97 #2 of 4 Archive leaderboard report
Referring Image Matting (Expression-based) RefMatte CLIPMat (ViT-B/16) SAD(E) 50.84 #2 of 4 Archive leaderboard report
Referring Image Matting (Keyword-based) RefMatte CLIPMat (ViT-L/14) MAD 0.0049 #1 of 4 Archive leaderboard report
Referring Image Matting (Keyword-based) RefMatte CLIPMat (ViT-L/14) MAD(E) 0.0051 #1 of 4 Archive leaderboard report
Referring Image Matting (Keyword-based) RefMatte CLIPMat (ViT-L/14) MSE 0.0022 #1 of 4 Archive leaderboard report
Referring Image Matting (Keyword-based) RefMatte CLIPMat (ViT-L/14) MSE(E) 0.0023 #1 of 4 Archive leaderboard report
Referring Image Matting (Keyword-based) RefMatte CLIPMat (ViT-L/14) SAD 8.51 #1 of 4 Archive leaderboard report
Referring Image Matting (Keyword-based) RefMatte CLIPMat (ViT-L/14) SAD(E) 8.98 #1 of 4 Archive leaderboard report
Referring Image Matting (Keyword-based) RefMatte CLIPMat (ViT-B/16) MAD 0.0057 #2 of 4 Archive leaderboard report
Referring Image Matting (Keyword-based) RefMatte CLIPMat (ViT-B/16) MAD(E) 0.0059 #2 of 4 Archive leaderboard report
Referring Image Matting (Keyword-based) RefMatte CLIPMat (ViT-B/16) MSE 0.0028 #2 of 4 Archive leaderboard report
Referring Image Matting (Keyword-based) RefMatte CLIPMat (ViT-B/16) MSE(E) 0.0029 #2 of 4 Archive leaderboard report
Referring Image Matting (Keyword-based) RefMatte CLIPMat (ViT-B/16) SAD 9.91 #2 of 4 Archive leaderboard report
Referring Image Matting (Keyword-based) RefMatte CLIPMat (ViT-B/16) SAD(E) 10.41 #2 of 4 Archive leaderboard report
Referring Image Matting (RefMatte-RW100) RefMatte CLIPMat (ViT-L/14) MAD 0.0510 #1 of 4 Archive leaderboard report
Referring Image Matting (RefMatte-RW100) RefMatte CLIPMat (ViT-L/14) MAD(E) 0.0505 #1 of 4 Archive leaderboard report
Referring Image Matting (RefMatte-RW100) RefMatte CLIPMat (ViT-L/14) MSE 0.0488 #1 of 4 Archive leaderboard report
Referring Image Matting (RefMatte-RW100) RefMatte CLIPMat (ViT-L/14) MSE(E) 0.0483 #1 of 4 Archive leaderboard report
Referring Image Matting (RefMatte-RW100) RefMatte CLIPMat (ViT-L/14) SAD 88.52 #1 of 4 Archive leaderboard report
Referring Image Matting (RefMatte-RW100) RefMatte CLIPMat (ViT-L/14) SAD(E) 87.92 #1 of 4 Archive leaderboard report
Referring Image Matting (RefMatte-RW100) RefMatte CLIPMat (ViT-B/16) MAD 0.0636 #2 of 4 Archive leaderboard report
Referring Image Matting (RefMatte-RW100) RefMatte CLIPMat (ViT-B/16) MAD(E) 0.0635 #2 of 4 Archive leaderboard report
Referring Image Matting (RefMatte-RW100) RefMatte CLIPMat (ViT-B/16) MSE 0.0614 #2 of 4 Archive leaderboard report
Referring Image Matting (RefMatte-RW100) RefMatte CLIPMat (ViT-B/16) MSE(E) 0.0612 #2 of 4 Archive leaderboard report
Referring Image Matting (RefMatte-RW100) RefMatte CLIPMat (ViT-B/16) SAD 110.66 #2 of 4 Archive leaderboard report
Referring Image Matting (RefMatte-RW100) RefMatte CLIPMat (ViT-B/16) SAD(E) 110.63 #2 of 4 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

Test

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections