Add model card for GemDepth

README.md

@@ -1,3 +1,59 @@
----
-license: apache-2.0
----
+---
+license: apache-2.0
+pipeline_tag: depth-estimation
+---
+
+# GemDepth: Geometry-Embedded Features for 3D-Consistent Video Depth
+
+GemDepth is a framework for 3D-consistent video depth estimation. It introduces a Geometry-Embedding Module (GEM) that predicts inter-frame camera poses to generate implicit geometric embeddings, providing the network with intrinsic 3D perception. Combined with an Alternating Spatio-Temporal Transformer (ASTT), GemDepth captures point-level correspondences to enhance spatial precision and enforce rigorous temporal consistency across complex dynamic scenarios.
+
+- **Paper:** [GemDepth: Geometry-Embedded Features for 3D-Consistent Video Depth](https://arxiv.org/abs/2605.10525)
+- **Repository:** [https://github.com/Yuecheng919/GemDepth](https://github.com/Yuecheng919/GemDepth)
+
+## Usage
+
+To use GemDepth, first clone the repository and install the requirements:
+
+```bash
+git clone https://github.com/Yuechengliu919/GemDepth
+cd GemDepth
+conda create -n gemdepth python=3.10
+conda activate gemdepth
+pip install -r requirements.txt
+```
+
+Then, you can initialize the model and perform inference using the following snippet:
+
+```python
+import torch
+from model.gemdepth import GemDepth
+
+DEVICE = 'cuda' if torch.cuda.is_available() else 'cpu'
+
+# Model configurations
+model_configs = {
+    'vits': {'encoder': 'vits', 'features': 64, 'out_channels': [48, 96, 192, 384]},
+    'vitl': {'encoder': 'vitl', 'features': 256, 'out_channels': [256, 512, 1024, 1024]},
+}
+
+# Initialize model (e.g., vitl)
+gemdepth = GemDepth(**model_configs['vitl'])
+checkpoint = torch.load("./checkpoint/gemdepth.pth", map_location='cpu', weights_only=False)
+gemdepth.load_state_dict(checkpoint, strict=True)
+gemdepth = gemdepth.to(DEVICE).eval()
+
+# Inference example
+# frames, target_fps = read_video_frames(video_path, max_len, target_fps, 1280)
+# depths, fps = gemdepth.infer_video_depth(frames, target_fps, input_size=1280, device=DEVICE)
+```
+
+## Citation
+
+```bibtex
+@article{liu2026gemdepth,
+  title={GemDepth: Geometry-Embedded Features for 3D-Consistent Video Depth},
+  author={Yuecheng Liu and others},
+  journal={arXiv preprint arXiv:2605.10525},
+  year={2026}
+}
+```