Research map / Vision-language and multimodal / CLIP and zero-shot
PixCLIP: Towards Fine-grained Vision-Language Understanding via Any-granularity Pixel-Text Alignment
YiCheng Xiao, Yu Chen, Hao-Xuan Ma, Jiale Hong, Caorui Li, Lingxiang Wu, Haiyun Guo, Jinqiao Wang
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 |
|---|---|---|---|
| add local-global representation enhancement loss L_LG (ablation) | PixCLIP (L_CL baseline) · RefCOCO Val | accuracy: 51.144% -> 59.041% | Improved |
| add fine-grained cropping alignment loss L_FC (ablation) | PixCLIP (L_CL + L_LG) · Urban1k I2T | Recall@1: 88.9% -> 91.0% | Improved |
| add L_FC to L_CL + L_LG on Ref-SAV M2T (ablation) | PixCLIP · Ref-SAV M2T | Recall@1: 47.4% -> 47.3% | No clear effect |
| add L_LG and L_FC to mask-text contrastive training (ablation) | PixCLIP · Urban1k T2I | Recall@1: 92.4% -> 93.7% | Improved |
| fuse mask embedding by adding to image patch embedding instead of concatenation (ablation) | PixCLIP · DOCCI I2T | Recall@1: 86.6 -> 96.7 | Improved |
| use full-image ratio 0.1 rather than 0.2 or 0.3 (ablation) | PixCLIP · Instance-COCO | top-1 accuracy: 72.0 (0.3) -> 72.7 (0.1) | Improved |
| train PixCLIP with mask prompts instead of Alpha-CLIP | PixCLIP · Ref-SAV M2T | Recall@1: 28.3% -> 47.3% | Improved |
| use PixCLIP versus LLM2CLIP on short-text retrieval | PixCLIP · Flickr30k I2T | Recall@1: 88.5 -> 84.2 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 on most shared results.
- Alpha-CLIP — Alpha-CLIP: A CLIP Model Focusing on Wherever You Want (CVPR 2024) · better on 6 of 6 shared results
- FG-CLIP — FG-CLIP: Fine-Grained Visual and Textual Alignment (ICML 2025) · better on 1 of 2 shared results
- MaskCLIP — MaskCLIP: Masked Self-Distillation Advances Contrastive Language-Image Pretraining (CVPR 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- FineLIP: Extending CLIP's Reach via Fine-Grained Alignment with Longer Text Inputs (CVPR 2025 · similarity 0.90)
- Aligning Forest and Trees in Images & Long Captions for Visually Grounded Understanding (NeurIPS 2026 · similarity 0.90)
- GOAL: Global-local Object Alignment Learning (CVPR 2025 · similarity 0.90)
- SmartCLIP: Modular Vision-language Alignment with Identification Guarantees (CVPR 2025 · similarity 0.90)
- SuperCLIP: CLIP with Simple Classification Supervision (NeurIPS 2025 · similarity 0.89)
- Visual-Text Cross Alignment: Refining the Similarity Score in Vision-Language Models (ICML 2024 · similarity 0.89)
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.
