Add comprehensive README for ViStream model checkpoint
Browse files
README.md
CHANGED
|
@@ -1,3 +1,50 @@
|
|
| 1 |
-
|
| 2 |
-
|
| 3 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# ViStream Model Checkpoint
|
| 2 |
+
|
| 3 |
+
This repository hosts the model checkpoint for **ViStream: Improving Computation Efficiency of Visual Streaming Perception via Law-of-Charge-Conservation Inspired Spiking Neural Network** (CVPR 2025).
|
| 4 |
+
|
| 5 |
+
## Model Description
|
| 6 |
+
|
| 7 |
+
ViStream is a novel framework that leverages the Law of Charge Conservation (LoCC) property in ST-BIF neurons and a differential encoding (DiffEncode) scheme to optimize SNN inference for Visual Streaming Perception. The framework achieves significant computational reduction while maintaining accuracy equivalent to its ANN counterpart across diverse VSP tasks including object detection, tracking, and segmentation.
|
| 8 |
+
|
| 9 |
+
## Repository Contents
|
| 10 |
+
|
| 11 |
+
- `checkpoint-90.pth` (292MB) - Pre-trained ViStream model checkpoint
|
| 12 |
+
|
| 13 |
+
## Usage
|
| 14 |
+
|
| 15 |
+
Download the checkpoint file and place it in your project directory:
|
| 16 |
+
|
| 17 |
+
```python
|
| 18 |
+
from huggingface_hub import hf_hub_download
|
| 19 |
+
|
| 20 |
+
# Download the checkpoint
|
| 21 |
+
checkpoint_path = hf_hub_download(
|
| 22 |
+
repo_id="AndyBlocker/ViStream",
|
| 23 |
+
filename="checkpoint-90.pth"
|
| 24 |
+
)
|
| 25 |
+
```
|
| 26 |
+
|
| 27 |
+
## Full Implementation
|
| 28 |
+
|
| 29 |
+
The complete ViStream implementation, demo videos, and documentation are available at:
|
| 30 |
+
**🔗 [GitHub Repository](https://github.com/Intelligent-Computing-Research-Group/ViStream)**
|
| 31 |
+
|
| 32 |
+
## Citation
|
| 33 |
+
|
| 34 |
+
```bibtex
|
| 35 |
+
@inproceedings{you2025vistream,
|
| 36 |
+
title={VISTREAM: Improving Computation Efficiency of Visual Streaming Perception via Law-of-Charge-Conservation Inspired Spiking Neural Network},
|
| 37 |
+
author={You, Kang and Wei, Ziling and Yan, Jing and Zhang, Boning and Guo, Qinghai and Zhang, Yaoyu and He, Zhezhi},
|
| 38 |
+
booktitle={Proceedings of the Computer Vision and Pattern Recognition Conference},
|
| 39 |
+
pages={8796--8805},
|
| 40 |
+
year={2025}
|
| 41 |
+
}
|
| 42 |
+
```
|
| 43 |
+
|
| 44 |
+
## Paper
|
| 45 |
+
|
| 46 |
+
📄 **[Read the full paper](https://openaccess.thecvf.com/content/CVPR2025/papers/You_VISTREAM_Improving_Computation_Efficiency_of_Visual_Streaming_Perception_via_Law-of-Charge-Conservation_CVPR_2025_paper.pdf)**
|
| 47 |
+
|
| 48 |
+
## License
|
| 49 |
+
|
| 50 |
+
This model is released under CC-BY-4.0 license.
|