Research map / Vision-language and multimodal / Multimodal LLMs
Benchmarking Omni-Vision Representation through the Lens of Visual Realms
Yuanhan Zhang, Zhenfei Yin, Jing Shao, Ziwei Liu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use Swin-T instead of ResNet-50 | ImageNet-1K supervised · OmniBenchmark | avg top-1: 34.3 -> 46.4 | Improved |
| pre-train ResNet-50 on 69M GV-D data | ResNet-50 · OmniBenchmark | avg top-1: 34.3 -> 45.4 | Improved |
| use self-supervised DINO instead of supervised RN50 | ResNet-50 · OmniBenchmark | avg top-1: 34.3 -> 38.9 | Improved |
| use MoCov2 instead of supervised RN50 | ResNet-50 · OmniBenchmark | avg top-1: 34.3 -> 34.8 | No clear effect |
| use SwAV-Places pre-training instead of ImageNet | ResNet-50 · OmniBenchmark | avg top-1: 34.3 -> 33.1 | Hurt |
| use masked image modeling (MAE) instead of supervised RN50 | ViT/ResNet-50 comparison · OmniBenchmark | avg top-1: 34.3 -> 30.6 | Hurt |
| use CutMix augmentation on ResNet-50 | ResNet-50 · OmniBenchmark | avg top-1: 34.3 -> 31.1 | Hurt |
| use ReLabel on ResNet-50 | ResNet-50 · OmniBenchmark | avg top-1: 34.3 -> 30.8 | Hurt |
| use MEAL-V2 distillation on ResNet-50 | ResNet-50 · OmniBenchmark | avg top-1: 34.3 -> 36.6 | Improved |
| use PeCo RN50 instead of RN50 | ResNet-50 · OmniBenchmark | avg top-1: 34.3 -> 36.4 | Improved |
| use ReCo RN101 instead of ResNet-101 | ResNet-101 · OmniBenchmark | avg top-1: 37.4 -> 38.7 | Improved |
| use CLIP image-text pre-training | ResNet-50 · OmniBenchmark | avg top-1: 34.3 -> 42.1 | Improved |
Closest papers by meaning
- Camouflaged Object Segmentation with Omni Perception (IJCV 2023 · similarity 0.84)
- Omnivore: A Single Model for Many Visual Modalities (CVPR 2022 · similarity 0.84)
- Omni-Captioner: Data Pipeline, Models, and Benchmark for Omni Detailed Perception (ICLR 2026 · similarity 0.81)
- HoVLE: Unleashing the Power of Monolithic Vision-Language Models with Holistic Vision-Language Embedding (CVPR 2025 · similarity 0.80)
- Omni-Scene: Omni-Gaussian Representation for Ego-Centric Sparse-View Scene Reconstruction (CVPR 2025 · similarity 0.80)
- ODI-Bench: Can MLLMs Understand Immersive Omnidirectional Environments? (ICLR 2026 · similarity 0.80)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
