update README

Author: Sharpiless

README.md

@@ -7,7 +7,7 @@
 
 ## 🧠 Overview
 
-**Lift to Match (L2M)** is a novel two-stage framework for **dense feature matching** that lifts 2D images into 3D space to enhance feature generalization and robustness. Unlike traditional methods that depend on multi-view image pairs, L2M is trained on large-scale, diverse single-view image collections.
+**Lift to Match (L2M)** is a two-stage framework for **dense feature matching** that lifts 2D images into 3D space to enhance feature generalization and robustness. Unlike traditional methods that depend on multi-view image pairs, L2M is trained on large-scale, diverse single-view image collections.
 
 - **Stage 1:** Learn a **3D-aware ViT-based encoder** using multi-view image synthesis and 3D Gaussian feature representation.
 - **Stage 2:** Learn a **feature decoder** through novel-view rendering and synthetic data, enabling robust matching across diverse scenarios.
@@ -38,51 +38,21 @@ Below are feature comparison results on the Sacré-Cœur dataset:
 
 ---
 
-## 🏗️ Data Generation (WIP)
-
-We synthesize multi-view images and 3D-aware Gaussian features from single-view inputs.  
-Scripts for data generation will be released soon.
-
----
-
-## 🏋️‍♀️ Model Training (Stage 1)
-
-We provide pretrained weights for the 3D-aware ViT encoder.
-
-> 🔗 **[Download pretrained encoder weights](#)** (Coming soon)
-
-You can visualize features using:
-
-```bash
-python vis_feats.py --input ./assets/sacre_coeur_A.jpg --model vit_encoder.pth
+To get the results, make sure your checkpoints and image files are in the correct paths, then run:
 ```
-
----
-
-## 🚀 Inference & Stage 2 (Coming Soon)
-
-The second stage—feature decoding with novel-view rendering—is **under development**. Stay tuned!
-
----
-
-## 📌 Citation
-
-```bibtex
-@article{liang2025lift2match,
-  title={Learning Dense Feature Matching via Lifting Single 2D Image to 3D Space},
-  author={Liang, Yingping and Hu, Yutao and Shao, Wenqi and Fu, Ying},
-  journal={ICCV},
-  year={2025}
-}
+python vis_feats.py \
+  --img_paths assets/sacre_coeur_A.jpg assets/sacre_coeur_B.jpg \
+  --ckpt_dino ckpts/dinov2.pth \
+  --ckpt_fit3d ckpts/fit3d.pth \
+  --ckpt_L2M ckpts/output_20250629/l2m_vit_base.pth \
+  --save_dir outputs_vis_feat
 ```
 
----
-
-## 📋 License
+## 🏗️ Data Generation
 
-This project is licensed under **CC BY 4.0**.
+We synthesize novel-view images with dense matching labels from single-view inputs for training.  
+Scripts for data generation will be released soon.
 
----
 
 ## 🙋‍♂️ Acknowledgements