Upload README.md

README.md

@@ -15,18 +15,13 @@ tags:
   - image-segmentation
 ---
 
-# PixDLM: A Dual-Path Multimodal Language Model for UAV Reasoning Segmentation
+# 【CVPR2026 Highlight】PixDLM: A Dual-Path Multimodal Language Model for UAV Reasoning Segmentation
 
 [![Paper](https://img.shields.io/badge/arXiv-2604.15670-b31b1b)](https://arxiv.org/abs/2604.15670)
 [![Model](https://img.shields.io/badge/HuggingFace-WhynotHug%2FPixDLM-yellow)](https://huggingface.co/WhynotHug/PixDLM)
 [![Dataset](https://img.shields.io/badge/HuggingFace-WhynotHug%2FDRSeg-green)](https://huggingface.co/datasets/WhynotHug/DRSeg)
+[![Project Page](https://img.shields.io/badge/Project-PixDLM-blue)](https://huggingface.co/spaces/WhynotHug/PixDLM)
 
-PixDLM is a dual-path multimodal language model for UAV reasoning segmentation.
-Given an aerial image and a reasoning-oriented question, PixDLM produces a target
-segmentation mask aligned with the described object.
-
-This repository contains the clean open-source release for the CVPR 2026
-Highlight paper and the 2027 CVPR Compute Transparency Champion release.
 
 ## Highlights
 
@@ -48,7 +43,7 @@ PixDLM/
 ├── docs/                     # Reproducibility and data/model documentation
 ├── examples/                 # Minimal local examples and expected file layout
 ├── pretrained/pixdlm-7b/     # Lightweight model config/tokenizer files
-├── release/huggingface/      # Model card, dataset card, and upload helpers
+├── release/huggingface/      # Model card, dataset card, Space, upload helpers
 ├── eval.py                   # Evaluation entry point
 ├── train_ds.py               # Training entry point
 └── requirements.txt
@@ -159,28 +154,6 @@ logs/<exp_name>/
 Typical per-sample artifacts include the input image, predicted mask, ground
 truth mask, red/green overlay, and a JSON result file.
 
-## DRSeg Evaluation Results
-
-The paper reports the following PixDLM metrics on DRSeg:
-
-| Reasoning type | gIoU | cIoU |
-| --- | ---: | ---: |
-| Attribute | 62.80 | 62.84 |
-| Scene | 61.75 | 64.03 |
-| Spatial | 62.51 | 62.80 |
-
-A practical 8-GPU verification run with this release produced:
-
-| Reasoning type | gIoU | cIoU |
-| --- | ---: | ---: |
-| Attribute | 62.51 | 60.82 |
-| Scene | 62.56 | 62.36 |
-| Spatial | 61.54 | 61.00 |
-| Overall | 62.20 | 61.41 |
-
-The gIoU reproduction is close to the paper table. The cIoU numbers can vary with
-the exact evaluation protocol, sampler behavior, and dependency versions.
-
 ## Dataset
 
 DRSeg is available at: