CogCanvas: A Benchmark for Evaluating Multi-Subject Reference-Based Image Generation

• URL: http://arxiv.org/abs/2606.15867v1
• Date: Sun, 14 Jun 2026 15:38:24 GMT
• Title: CogCanvas: A Benchmark for Evaluating Multi-Subject Reference-Based Image Generation
• Abstract summary: We introduce CogCanvas, a benchmark of 1,952 curated reference images spanning 100 celebrity identities, 115 distinctive objects and fashion items, and 29 real-world background scenes.<n>The pipeline combines DINOv2-based deduplication, two-stage aesthetic filtering, and automated derivation of structured interaction and position graphs.<n> Benchmarking five SOTA methods reveals that every model degrades substantially as group size grows from 2 to 5, with near-complete failure on object/fashion binding beyond three subjects.
• Score: 19.10912769596512
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Multi-subject reference-based image generation requires jointly preserving multiple human identities, binding per-person objects and fashion items, and respecting a specified background scene, a regime where current diffusion models remain brittle. Existing benchmarks evaluate only one axis at a time and none jointly captures multi-identity composition with human-object interaction, background grounding, and spatial plausibility. We introduce CogCanvas, a benchmark of 1,952 curated reference images spanning 100 celebrity identities, 115 distinctive objects and fashion items, and 29 real-world background scenes including landmarks, from which we construct 1,361 compositional prompts covering 2-5 person group sizes. The curation pipeline combines DINOv2-based deduplication, two-stage aesthetic filtering, and automated derivation of structured interaction and position graphs that serve as ground-truth supervision. CogCanvas supports three tasks, reference-based multi-human-object generation (primary), text-to-image compositional generation, and reference retrieval, under a unified six-axis evaluation protocol. We introduce two metrics tailored to the multi-reference setting: BG-Sim, which scores background fidelity on SAM 3-masked regions via DINOv3 feature similarity, and Attr-VQA, which uses a multimodal LLM to verify per-subject attribute binding and inter-person interactions against the structured graphs. Benchmarking five SOTA methods reveals that every model degrades substantially as group size grows from 2 to 5, with near-complete failure on object/fashion binding beyond three subjects.

Related papers

• 3SGen: Unified Subject, Style, and Structure-Driven Image Generation with Adaptive Task-specific Memory [54.056509629389915]
  3SGen is a task-aware unified framework that performs all three conditioning modes within a single model.<n>At its core, an Adaptive Task-specific Memory (ATM) module dynamically disentangles, stores, and retrieves condition-specific priors.<n>We propose 3SGen-Bench, a unified image-driven generation benchmark with standardized metrics for evaluating cross-task fidelity and controllability.
  arXiv  Detail & Related papers  (2025-12-22T11:07:27Z)
• TokenAR: Multiple Subject Generation via Autoregressive Token-level enhancement [87.82338951215131]
  TokenAR is a simple but effective token-level enhancement mechanism to address reference identity confusion problem.<n>Instruct Token Injection plays as a role of extra visual feature container to inject detailed and complementary priors for reference tokens.<n>The identity-token disentanglement strategy (ITD) explicitly guides the token representations toward independently representing the features of each identity.
  arXiv  Detail & Related papers  (2025-10-18T03:36:26Z)
• SegMASt3R: Geometry Grounded Segment Matching [23.257530861472656]
  We leverage the spatial understanding of 3D foundation models to tackle wide-baseline segment matching.<n>We propose an architecture that uses the inductive bias of these 3D foundation models to match segments across image pairs with up to 180 degree view-point change rotation.
  arXiv  Detail & Related papers  (2025-10-06T17:31:32Z)
• DIPO: Dual-State Images Controlled Articulated Object Generation Powered by Diverse Data [67.99373622902827]
  DIPO is a framework for controllable generation of articulated 3D objects from a pair of images.<n>We propose a dual-image diffusion model that captures relationships between the image pair to generate part layouts and joint parameters.<n>We propose PM-X, a large-scale dataset of complex articulated 3D objects, accompanied by rendered images, URDF annotations, and textual descriptions.
  arXiv  Detail & Related papers  (2025-05-26T18:55:14Z)
• MMIG-Bench: Towards Comprehensive and Explainable Evaluation of Multi-Modal Image Generation Models [42.91502354577658]
  MMIG-Bench is a comprehensive Multi-Modal Image Generation Benchmark.<n>It pairs 4,850 richly annotated text prompts with 1,750 multi-view reference images across 380 subjects.<n>Using MMIG-Bench, we benchmark 17 state-of-the-art models, including Gemini 2.5 Pro, FLUX, DreamBooth, and IP-Adapter.
  arXiv  Detail & Related papers  (2025-05-26T02:07:24Z)
• Context-Aware Aerial Object Detection: Leveraging Inter-Object and Background Relationships [10.416444779297148]
  In most modern object detection pipelines, the detection proposals are processed independently given the feature map.<n>We propose a framework that leverages the strengths of Transformer-based models and Contrastive Language-Image Pre-training features.<n>Our approach achieves consistent improvements, setting new state-of-the-art results with increases of 1.37 mAP$_50$ on DOTA-v1.0, 5.30 mAP$_50$ on DOTA-v1.5, 2.30 mAP$_50$ on DOTA-v2.0 and 3.23 mAP$_50$ on DI
  arXiv  Detail & Related papers  (2024-04-05T14:39:13Z)
• SM$^3$: Self-Supervised Multi-task Modeling with Multi-view 2D Images for Articulated Objects [24.737865259695006]
  We propose a self-supervised interaction perception method, referred to as SM$3$, to model articulated objects. By constructing 3D geometries and textures from the captured 2D images, SM$3$ achieves integrated optimization of movable part and joint parameters. Evaluations demonstrate that SM$3$ surpasses existing benchmarks across various categories and objects, while its adaptability in real-world scenarios has been thoroughly validated.
  arXiv  Detail & Related papers  (2024-01-17T11:15:09Z)
• Collaborative Group: Composed Image Retrieval via Consensus Learning from Noisy Annotations [67.92679668612858]
  We propose the Consensus Network (Css-Net), inspired by the psychological concept that groups outperform individuals. Css-Net comprises two core components: (1) a consensus module with four diverse compositors, each generating distinct image-text embeddings; and (2) a Kullback-Leibler divergence loss that encourages learning of inter-compositor interactions. On benchmark datasets, particularly FashionIQ, Css-Net demonstrates marked improvements. Notably, it achieves significant recall gains, with a 2.77% increase in R@10 and 6.67% boost in R@50, underscoring its
  arXiv  Detail & Related papers  (2023-06-03T11:50:44Z)
• T-Person-GAN: Text-to-Person Image Generation with Identity-Consistency and Manifold Mix-Up [16.165889084870116]
  We present an end-to-end approach to generate high-resolution person images conditioned on texts only. We develop an effective generative model to produce person images with two novel mechanisms.
  arXiv  Detail & Related papers  (2022-08-18T07:41:02Z)
• Graph-Based 3D Multi-Person Pose Estimation Using Multi-View Images [79.70127290464514]
  We decompose the task into two stages, i.e. person localization and pose estimation. And we propose three task-specific graph neural networks for effective message passing. Our approach achieves state-of-the-art performance on CMU Panoptic and Shelf datasets.
  arXiv  Detail & Related papers  (2021-09-13T11:44:07Z)
• Object-Centric Image Generation from Layouts [93.10217725729468]
  We develop a layout-to-image-generation method to generate complex scenes with multiple objects. Our method learns representations of the spatial relationships between objects in the scene, which lead to our model's improved layout-fidelity. We introduce SceneFID, an object-centric adaptation of the popular Fr'echet Inception Distance metric, that is better suited for multi-object images.
  arXiv  Detail & Related papers  (2020-03-16T21:40:09Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.