Add tag
#1
by
nielsr
HF Staff
- opened
README.md
CHANGED
|
@@ -1,84 +1,85 @@
|
|
| 1 |
-
---
|
| 2 |
-
license: mit
|
| 3 |
-
library_name: BEVANet
|
| 4 |
-
tags:
|
| 5 |
-
-
|
| 6 |
-
- real-time
|
| 7 |
-
-
|
| 8 |
-
-
|
| 9 |
-
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
|
| 13 |
-
|
| 14 |
-
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
<
|
| 20 |
-
|
| 21 |
-
|
| 22 |
-
[](https://arxiv.org/abs/2508.07300)
|
| 23 |
+
[](https://ieeexplore.ieee.org/document/11084676)
|
| 24 |
+
[](https://github.com/maomao0819/BEVANet)
|
| 25 |
+
[](https://opensource.org/licenses/MIT)
|
| 26 |
+
|
| 27 |
+
[Ping-Mao Huang](https://www.linkedin.com/in/maomao0819/), I-Tien Chao, Ping-Chia Huang, Jia-Wei Liao, Yung-Yu Chuang<br>
|
| 28 |
+
National Taiwan University
|
| 29 |
+
|
| 30 |
+
</div>
|
| 31 |
+
|
| 32 |
+
## Model Description
|
| 33 |
+
|
| 34 |
+
- **Task**: Real-Time Semantic Segmentation
|
| 35 |
+
- **Dataset**: Cityscapes
|
| 36 |
+
- **Model**: BEVANet
|
| 37 |
+
|
| 38 |
+
BEVANet semantic segmentation model trained on Cityscapes dataset. Achieves 81.0% mIoU with 32.8 FPS on RTX3090.
|
| 39 |
+
|
| 40 |
+
## Performance
|
| 41 |
+
|
| 42 |
+
- **mIoU**: 81.0%
|
| 43 |
+
- **FPS**: 32.8 (RTX3090)
|
| 44 |
+
- **Parameters**: 58.6M
|
| 45 |
+
|
| 46 |
+
## Usage
|
| 47 |
+
|
| 48 |
+
```python
|
| 49 |
+
from models.BEVANet import BEVANet_SEG
|
| 50 |
+
|
| 51 |
+
# Load model from specific branch
|
| 52 |
+
model = BEVANet_SEG.from_pretrained(
|
| 53 |
+
"maomao0819/BEVANet",
|
| 54 |
+
revision="main"
|
| 55 |
+
)
|
| 56 |
+
|
| 57 |
+
# For main branch (no revision needed)
|
| 58 |
+
# model = BEVANet_SEG.from_pretrained("maomao0819/BEVANet")
|
| 59 |
+
```
|
| 60 |
+
|
| 61 |
+
## All Model Variants
|
| 62 |
+
|
| 63 |
+
| Model | Branch | Dataset | Task | Performance | FPS | Parameters |
|
| 64 |
+
|-------|--------|---------|------|-------------|-----|------------|
|
| 65 |
+
| BEVANet-S | `imagenet-bevanet-s` | ImageNet | Classification | 71.1% Top-1 | 198.6 | 16.3M |
|
| 66 |
+
| BEVANet | `imagenet-bevanet` | ImageNet | Classification | 76.3% Top-1 | 82.3 | 57.4M |
|
| 67 |
+
| BEVANet-S | `cityscapes-bevanet-s` | Cityscapes | Segmentation | 78.2% mIoU | 70.0 | 15.2M |
|
| 68 |
+
| BEVANet | `main` | Cityscapes | Segmentation | 81.0% mIoU | 32.8 | 58.6M |
|
| 69 |
+
| BEVANet-S | `camvid-bevanet-s` | CamVid | Segmentation | 83.1% mIoU | 79.4 | 15.2M |
|
| 70 |
+
| BEVANet | `ade20k-bevanet` | ADE20K | Segmentation | 39.8% mIoU | 73.3 | 58.9M |
|
| 71 |
+
|
| 72 |
+
## Citation
|
| 73 |
+
|
| 74 |
+
If you use this model, please cite:
|
| 75 |
+
|
| 76 |
+
```bibtex
|
| 77 |
+
@inproceedings{huang2025bevanet,
|
| 78 |
+
title={Bevanet: Bilateral Efficient Visual Attention Network for Real-Time Semantic Segmentation},
|
| 79 |
+
author={Huang, Ping-Mao and Chao, I-Tien and Huang, Ping-Chia and Liao, Jia-Wei and Chuang, Yung-Yu},
|
| 80 |
+
booktitle={2025 IEEE International Conference on Image Processing (ICIP)},
|
| 81 |
+
pages={2778--2783},
|
| 82 |
+
year={2025},
|
| 83 |
+
organization={IEEE}
|
| 84 |
+
}
|
| 85 |
+
```
|