update
Browse files
README.md
CHANGED
|
@@ -16,6 +16,7 @@ license: apache-2.0
|
|
| 16 |
>[From Static to Dynamic: Adapting Landmark-Aware Image Models for Facial Expression Recognition in Videos](https://ieeexplore.ieee.org/document/10663980)<br>
|
| 17 |
>Yin Chen$^{β }$, Jia Li$^{β β}$, Shiguang Shan, Meng Wang, and Richang Hong
|
| 18 |
|
|
|
|
| 19 |
## π° News
|
| 20 |
**[2024.9.5]** The fine-tuned checkpoints are available.
|
| 21 |
|
|
@@ -28,4 +29,31 @@ license: apache-2.0
|
|
| 28 |
## π Main Results
|
| 29 |
|
| 30 |
### Dynamic Facial Expression Recognition
|
| 31 |
-
<img width="1024" alt="image" src="https://github.com/user-attachments/assets/2144837d-9fd5-4f88-8447-1f6049b38e9a">
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 16 |
>[From Static to Dynamic: Adapting Landmark-Aware Image Models for Facial Expression Recognition in Videos](https://ieeexplore.ieee.org/document/10663980)<br>
|
| 17 |
>Yin Chen$^{β }$, Jia Li$^{β β}$, Shiguang Shan, Meng Wang, and Richang Hong
|
| 18 |
|
| 19 |
+
|
| 20 |
## π° News
|
| 21 |
**[2024.9.5]** The fine-tuned checkpoints are available.
|
| 22 |
|
|
|
|
| 29 |
## π Main Results
|
| 30 |
|
| 31 |
### Dynamic Facial Expression Recognition
|
| 32 |
+
<img width="1024" alt="image" src="https://github.com/user-attachments/assets/2144837d-9fd5-4f88-8447-1f6049b38e9a">
|
| 33 |
+
|
| 34 |
+
<img width="1024" alt="image" src="https://github.com/user-attachments/assets/4a80731e-666e-4cef-9f74-5f794eea7116">
|
| 35 |
+
|
| 36 |
+
|
| 37 |
+
### Static Facial Expression Recognition
|
| 38 |
+
<img width="1024" alt="image" src="https://github.com/user-attachments/assets/89a47ea3-1036-4124-927c-563af8007d1f">
|
| 39 |
+
|
| 40 |
+
### Visualization
|
| 41 |
+
<img width="1024" alt="image" src="https://github.com/user-attachments/assets/aea1385d-0d1b-4f5e-8775-087a30363751">
|
| 42 |
+
|
| 43 |
+
|
| 44 |
+
## Fine-tune with pre-trained weights
|
| 45 |
+
1γ Download the pre-trained weights from [baidu drive](https://pan.baidu.com/s/1J5eCnTn_Wpn0raZTIUCfgw?pwd=dji4) or [google drive](https://drive.google.com/file/d/1Y9zz8z_LwUi-tSFBAwDPZkVoyY6mhZlu/view?usp=drive_link) or [onedrive](https://mailhfuteducn-my.sharepoint.com/:f:/g/personal/2022111029_mail_hfut_edu_cn/EgKQNq8Y2chKl2TSoYf_OA0BQpCwx-FDw2ksPaMxBntZ8A), and move it to the ckpts directory.
|
| 46 |
+
|
| 47 |
+
2γ Run the following command to fine-tune the model on the target dataset.
|
| 48 |
+
```bash
|
| 49 |
+
conda create -n s2d python=3.9
|
| 50 |
+
conda activate s2d
|
| 51 |
+
pip install -r requirements.txt
|
| 52 |
+
bash run.sh
|
| 53 |
+
```
|
| 54 |
+
|
| 55 |
+
## π Reported Results and Fine-tuned Weights
|
| 56 |
+
The fine-tuned checkpoints can be downloaded from [baidu driver](https://pan.baidu.com/s/1Xz5j8QW32x7L0bnTEorUbA?pwd=5drk) or [huggingface](https://huggingface.co/cyinen/S2D).
|
| 57 |
+
<table border="1" cellspacing="0" cellpadding="5">
|
| 58 |
+
[](https://star-history.com/#MSA-LMC/S2D&Date)
|
| 59 |
+
|