|
|
--- |
|
|
license: cc-by-nc-4.0 |
|
|
language: |
|
|
- en |
|
|
metrics: |
|
|
- roc_auc |
|
|
- matthews_correlation |
|
|
pipeline_tag: video-classification |
|
|
--- |
|
|
# Simplifying Traffic Anomaly Detection with Video Foundation Models |
|
|
|
|
|
Svetlana Orlova, Tommie Kerssies, Brunó B. Englert, Gijs Dubbelman \ |
|
|
Eindhoven University of Technology |
|
|
|
|
|
[](https://arxiv.org/abs/2507.09338) |
|
|
[](https://huggingface.co/tue-mps/simple-tad/tree/main/models) |
|
|
[](https://github.com/tue-mps/simple-tad) |
|
|
|
|
|
<table> |
|
|
<tr> |
|
|
<td> |
|
|
<a href="https://youtu.be/hY2hUlTNhCU" target="_blank"> |
|
|
<img src="illustrations/videos/PYL3JcSsS6o_004036_small-dapt.gif" width="100%"> |
|
|
</a> |
|
|
</td> |
|
|
<td> |
|
|
<a href="https://youtu.be/tKe2nTIHf9k" target="_blank"> |
|
|
<img src="illustrations/videos/Sihe6aeyLHg_000602_small-dapt.gif" width="100%"> |
|
|
</a> |
|
|
</td> |
|
|
</tr> |
|
|
</table> |
|
|
|
|
|
Recent methods for ego-centric Traffic Anomaly Detection (TAD) often rely on complex multi-stage or multi-representation fusion architectures, yet it remains unclear whether such complexity is necessary. Recent findings in visual perception suggest that foundation models, enabled by advanced pre-training, allow simple yet flexible architectures to outperform specialized designs. Therefore, in this work, we investigate an architecturally simple encoder-only approach using plain Video Vision Transformers (Video ViTs) and study how pre-training enables strong TAD performance. We find that: (i) advanced pre-training enables simple encoder-only models to match or even surpass the performance of specialized state-of-the-art TAD methods, while also being significantly more efficient; (ii) although weakly- and fully-supervised pre-training are advantageous on standard benchmarks, we find them less effective for TAD. Instead, self-supervised Masked Video Modeling (MVM) provides the strongest signal; and (iii) Domain-Adaptive Pre-Training (DAPT) on unlabeled driving videos further improves downstream performance, without requiring anomalous examples. Our findings highlight the importance of pre-training and show that effective, efficient, and scalable TAD models can be built with minimal architectural complexity. |
|
|
|
|
|
 |
|
|
|
|
|
### ✨ DoTA and DADA-2000 results |
|
|
|
|
|
 |
|
|
|
|
|
Video ViT-based encoder-only models set a new state of the art |
|
|
on both datasets, while being significantly more efficient than top-performing specialized methods. FPS measured using NVIDIA A100 |
|
|
MIG, 2 1 GPU. † From prior work. ‡ Optimistic estimates using publicly available components of the model. “A→B”: trained on A, tested |
|
|
on B; D2K: DADA-2000. |
|
|
|
|
|
## 🧩 Code |
|
|
|
|
|
Check out our GitHub repo: **[simple-tad](https://github.com/tue-mps/simple-tad)** |
|
|
|
|
|
## 📍Model Zoo |
|
|
|
|
|
### DAPT (adapted) models |
|
|
| Method | Backbone | Initialized with | DAPT epochs | DAPT data | Checkpoint | |
|
|
| :------: |:--------:|:---------------:|:-----------:|:------------------:|:----------:| |
|
|
| VideoMAE | ViT-S | [Kinetics-400 1600 ep](https://github.com/MCG-NJU/VideoMAE/blob/main/MODEL_ZOO.md#kinetics-400) | 12 | Kinetics-700 | [simpletad_dapt-k700_videomae-s_ep12.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/DAPT/simpletad_dapt-k700_videomae-s_ep12.pth) | |
|
|
| VideoMAE | ViT-S | [Kinetics-400 1600 ep](https://github.com/MCG-NJU/VideoMAE/blob/main/MODEL_ZOO.md#kinetics-400) | 12 | BDD100K | [simpletad_dapt-onlybdd_videomae-s_ep12.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/DAPT/simpletad_dapt-onlybdd_videomae-s_ep12.pth) | |
|
|
| VideoMAE | ViT-S | [Kinetics-400 1600 ep](https://github.com/MCG-NJU/VideoMAE/blob/main/MODEL_ZOO.md#kinetics-400) | 12 | BDD100K + CAP-DATA | [simpletad_dapt_videomae-s_ep12.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/DAPT/simpletad_dapt_videomae-s_ep12.pth) | |
|
|
| VideoMAE | ViT-B | [Kinetics-400 1600 ep](https://github.com/MCG-NJU/VideoMAE/blob/main/MODEL_ZOO.md#kinetics-400) | 12 | BDD100K + CAP-DATA | [simpletad_dapt_videomae-b_ep12.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/DAPT/simpletad_dapt_videomae-b_ep12.pth) | |
|
|
| VideoMAE | ViT-L | [Kinetics-400 1600 ep](https://github.com/MCG-NJU/VideoMAE/blob/main/MODEL_ZOO.md#kinetics-400) | 12 | BDD100K + CAP-DATA | [simpletad_dapt_videomae-l_ep12.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/DAPT/simpletad_dapt_videomae-l_ep12.pth) | |
|
|
|
|
|
|
|
|
### Fine-tuned on DoTA |
|
|
| Method | Backbone | Initialized with | Best AUC<sub>ROC</sub> checkpoint | Best AUC<sub>MCC</sub> checkpoint | AUC<sub>ROC</sub> | AUC<sub>MCC</sub> | |
|
|
|:---------:|:--------:|:--------------------------------:|:-----------------------:|:-----------------------:|:-------:|---------| |
|
|
| VideoMAE | ViT-S | VideoMAE ([Kinetics-400 1600 ep](https://github.com/MCG-NJU/VideoMAE/blob/main/MODEL_ZOO.md#kinetics-400)) | [simpletad_ft-dota_vm1-s_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_vm1-s_auroc.pth) | [simpletad_ft-dota_vm1-s_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_vm1-s_aumcc.pth) | 83.7 | 46.9 | |
|
|
| VideoMAE | ViT-B | VideoMAE ([Kinetics-400 1600 ep](https://github.com/MCG-NJU/VideoMAE/blob/main/MODEL_ZOO.md#kinetics-400)) | [simpletad_ft-dota_vm1-b-1600_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_vm1-b-1600_auroc.pth) | [simpletad_ft-dota_vm1-b-1600_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_vm1-b-1600_aumcc.pth) | 86.3 | 54.8 | |
|
|
| VideoMAE | ViT-L | VideoMAE ([Kinetics-400 1600 ep](https://github.com/MCG-NJU/VideoMAE/blob/main/MODEL_ZOO.md#kinetics-400)) | [simpletad_ft-dota_vm1-l_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_vm1-l_auroc.pth) | [simpletad_ft-dota_vm1-l_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_vm1-l_aumcc.pth) | 88.2 | 58.7 | |
|
|
| VideoMAE2 | ViT-S | VideoMAE2 ([vit_s_k710_dl_from_giant.pth](https://github.com/OpenGVLab/VideoMAEv2/blob/master/docs/MODEL_ZOO.md#distillation)) | [simpletad_ft-dota_vm2-s_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_vm2-s_auroc.pth) | [simpletad_ft-dota_vm2-s_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_vm2-s_aumcc.pth) | 86.0 | 54.1 | |
|
|
| VideoMAE2 | ViT-B | VideoMAE2 ([vit_b_k710_dl_from_giant.pth](https://github.com/OpenGVLab/VideoMAEv2/blob/master/docs/MODEL_ZOO.md#distillation)) | [simpletad_ft-dota_vm2-b_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_vm2-b_auroc.pth) | [simpletad_ft-dota_vm2-b_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_vm2-b_aumcc.pth) | 86.9 | 55.4 | |
|
|
| MVD_fromL | ViT-S | MVD ([Kinetics-400 Teacher ViT-L](https://github.com/ruiwang2021/mvd/blob/main/MODEL_ZOO.md#kinetics-400)) | [simpletad_ft-dota_mvd-s-fromL_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_mvd-s-fromL_auroc.pth) | [simpletad_ft-dota_mvd-s-fromL_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_mvd-s-fromL_aumcc.pth) | 85.3 | 53.8 | |
|
|
| MVD_fromB | ViT-B | MVD ([Kinetics-400 Teacher ViT-B](https://github.com/ruiwang2021/mvd/blob/main/MODEL_ZOO.md#kinetics-400)) | [simpletad_ft-dota_mvd-b-fromB_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_mvd-b-fromB_auroc.pth) | [simpletad_ft-dota_mvd-b-fromB_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_mvd-b-fromB_aumcc.pth) | 86.1 | 54.7 | |
|
|
| MVD_fromL | ViT-L | MVD ([Kinetics-400 Teacher ViT-L](https://github.com/ruiwang2021/mvd/blob/main/MODEL_ZOO.md#kinetics-400)) | [simpletad_ft-dota_mvd-l-fromL_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_mvd-l-fromL_auroc.pth) | [simpletad_ft-dota_mvd-l-fromL_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_mvd-l-fromL_aumcc.pth) | 87.2 | 58.1 | |
|
|
| **DAPT-** VideoMAE | ViT-S | DAPT (BDD100K + CAP-DATA) | [simpletad_ft-dota_dapt-vm1-s_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_dapt-vm1-s_auroc.pth) | [simpletad_ft-dota_dapt-vm1-s_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_dapt-vm1-s_aumcc.pth) | 86.4 | 54.0 | |
|
|
| **DAPT-** VideoMAE | ViT-B | DAPT (BDD100K + CAP-DATA) | [simpletad_ft-dota_dapt-vm1-b_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_dapt-vm1-b_auroc.pth) | [simpletad_ft-dota_dapt-vm1-b_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_dapt-vm1-b_aumcc.pth) | 87.9 | 57.5 | |
|
|
| **DAPT-** VideoMAE | ViT-L | DAPT (BDD100K + CAP-DATA) | [simpletad_ft-dota_dapt-vm1-l_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_dapt-vm1-l_auroc.pth) | [simpletad_ft-dota_dapt-vm1-l_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_DoTA/simpletad_ft-dota_dapt-vm1-l_aumcc.pth) | 88.4 | 58.9 | |
|
|
|
|
|
|
|
|
### Fine-tuned on DADA-2000 |
|
|
| Method | Backbone | Initialized with | Best AUC<sub>ROC</sub> checkpoint | Best AUC<sub>MCC</sub> checkpoint | AUC<sub>ROC</sub> | AUC<sub>MCC</sub> | |
|
|
|:---------:|:--------:|:--------------------------------:|:-----------------------:|:-----------------------:|:-------:|---------| |
|
|
| VideoMAE | ViT-S | VideoMAE ([Kinetics-400 1600 ep](https://github.com/MCG-NJU/VideoMAE/blob/main/MODEL_ZOO.md#kinetics-400)) | [simpletad_ft-dada_vm1-s_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_vm1-s_auroc.pth) | [simpletad_ft-dada_vm1-s_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_vm1-s_aumcc.pth) | 83.0 | 48.2 | |
|
|
| VideoMAE | ViT-B | VideoMAE ([Kinetics-400 1600 ep](https://github.com/MCG-NJU/VideoMAE/blob/main/MODEL_ZOO.md#kinetics-400)) | [simpletad_ft-dada_vm1-b-1600_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_vm1-b-1600_auroc.pth) | [simpletad_ft-dada_vm1-b-1600_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_vm1-b-1600_aumcc.pth) | 85.4 | 52.2 | |
|
|
| VideoMAE | ViT-L | VideoMAE ([Kinetics-400 1600 ep](https://github.com/MCG-NJU/VideoMAE/blob/main/MODEL_ZOO.md#kinetics-400)) | [simpletad_ft-dada_vm1-l_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_vm1-l_auroc.pth) | [simpletad_ft-dada_vm1-l_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_vm1-l_aumcc.pth) | 87.2 | 55.4 | |
|
|
| VideoMAE2 | ViT-S | VideoMAE2 ([vit_s_k710_dl_from_giant.pth](https://github.com/OpenGVLab/VideoMAEv2/blob/master/docs/MODEL_ZOO.md#distillation)) | [simpletad_ft-dada_vm2-s_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_vm2-s_auroc.pth) | [simpletad_ft-dada_vm2-s_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_vm2-s_aumcc.pth) | 84.8 | 50.3 | |
|
|
| VideoMAE2 | ViT-B | VideoMAE2 ([vit_b_k710_dl_from_giant.pth](https://github.com/OpenGVLab/VideoMAEv2/blob/master/docs/MODEL_ZOO.md#distillation)) | [simpletad_ft-dada_vm2-b_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_vm2-b_auroc.pth) | [simpletad_ft-dada_vm2-b_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_vm2-b_aumcc.pth) | 86.3 | 53.3 | |
|
|
| MVD_fromL | ViT-S | MVD ([Kinetics-400 Teacher ViT-L](https://github.com/ruiwang2021/mvd/blob/main/MODEL_ZOO.md#kinetics-400)) | [simpletad_ft-dada_mvd-s-fromL_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_mvd-s-fromL_auroc.pth) | [simpletad_ft-dada_mvd-s-fromL_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_mvd-s-fromL_aumcc.pth) | 82.2 | 50.2 | |
|
|
| MVD_fromB | ViT-B | MVD ([Kinetics-400 Teacher ViT-B](https://github.com/ruiwang2021/mvd/blob/main/MODEL_ZOO.md#kinetics-400)) | [simpletad_ft-dada_mvd-b-fromB_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_mvd-b-fromB_auroc.pth) | [simpletad_ft-dada_mvd-b-fromB_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_mvd-b-fromB_aumcc.pth) | 84.7 | 50.9 | |
|
|
| MVD_fromL | ViT-L | MVD ([Kinetics-400 Teacher ViT-L](https://github.com/ruiwang2021/mvd/blob/main/MODEL_ZOO.md#kinetics-400)) | [simpletad_ft-dada_mvd-l-fromL_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_mvd-l-fromL_auroc.pth) | [simpletad_ft-dada_mvd-l-fromL_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_mvd-l-fromL_aumcc.pth) | 86.1 | 53.7 | |
|
|
| **DAPT-** VideoMAE | ViT-S | DAPT (BDD100K + CAP-DATA) | [simpletad_ft-dada_dapt-vm1-s_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_dapt-vm1-s_auroc.pth) | [simpletad_ft-dada_dapt-vm1-s_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_dapt-vm1-s_aumcc.pth) | 85.6 | 52.0 | |
|
|
| **DAPT-** VideoMAE | ViT-B | DAPT (BDD100K + CAP-DATA) | [simpletad_ft-dada_dapt-vm1-b_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_dapt-vm1-b_auroc.pth) | [simpletad_ft-dada_dapt-vm1-b_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_dapt-vm1-b_aumcc.pth) | 87.6 | 55.2 | |
|
|
| **DAPT-** VideoMAE | ViT-L | DAPT (BDD100K + CAP-DATA) | [simpletad_ft-dada_dapt-vm1-l_auroc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_dapt-vm1-l_auroc.pth) | [simpletad_ft-dada_dapt-vm1-l_aumcc.pth](https://huggingface.co/tue-mps/simple-tad/resolve/main/models/Finetune_D2K/simpletad_ft-dada_dapt-vm1-l_aumcc.pth) | 88.5 | 56.8 | |
|
|
|
|
|
|
|
|
## ☎️ Contact |
|
|
|
|
|
Svetlana Orlova: s.orlova@tue.nl, orsveri@gmail.com |
|
|
|
|
|
## 👍 Acknowledgements |
|
|
|
|
|
Our code is mainly based on the [VideoMAE](https://github.com/MCG-NJU/VideoMAE) codebase. |
|
|
With Video ViTs that have identical architecture, we only used their weights: |
|
|
[ViViT](https://github.com/google-research/scenic/blob/main/scenic/projects/vivit/README.md), |
|
|
[VideoMAE2](https://github.com/OpenGVLab/VideoMAEv2), |
|
|
[SMILE](https://github.com/fmthoker/SMILE), |
|
|
[SIGMA](https://github.com/QUVA-Lab/SIGMA/), |
|
|
[MME](https://github.com/XinyuSun/MME), |
|
|
[MGMAE](https://github.com/MCG-NJU/MGMAE). \ |
|
|
We used fragments of original implementations of |
|
|
[MVD](https://github.com/ruiwang2021/mvd), |
|
|
[InternVideo2](https://github.com/OpenGVLab/InternVideo/tree/main/InternVideo2/single_modality), |
|
|
and [UMT](https://github.com/OpenGVLab/unmasked_teacher/tree/main/single_modality) to integrate these models with our codebase. |
|
|
|
|
|
## ✏️ Citation |
|
|
|
|
|
If you think this project is helpful, please feel free to like us ❤️ and cite our paper: |
|
|
|
|
|
``` |
|
|
@inproceedings{orlova2025simplifying, |
|
|
title={Simplifying Traffic Anomaly Detection with Video Foundation Models}, |
|
|
author={Orlova, Svetlana and Kerssies, Tommie and Englert, Brun{\'o} B and Dubbelman, Gijs}, |
|
|
booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision}, |
|
|
year={2025} |
|
|
} |
|
|
|
|
|
@article{orlova2025simplifying, |
|
|
title={Simplifying Traffic Anomaly Detection with Video Foundation Models}, |
|
|
author={Orlova, Svetlana and Kerssies, Tommie and Englert, Brun{\'o} B and Dubbelman, Gijs}, |
|
|
journal={arXiv preprint arXiv:2507.09338}, |
|
|
year={2025} |
|
|
} |
|
|
``` |
|
|
|
|
|
## |
|
|
|
|
|
<table> |
|
|
<tr> |
|
|
<td> |
|
|
<a href="https://youtu.be/vPBKj9SF9yg" target="_blank"> |
|
|
<img src="illustrations/videos/0RJPQ_97dcs_004503_large-dapt.gif" width="100%"> |
|
|
</a> |
|
|
</td> |
|
|
<td> |
|
|
<a href="https://youtu.be/vGaYPZEuv5k" target="_blank"> |
|
|
<img src="illustrations/videos/y4Evv5By6sg_004171_large-dapt.gif" width="100%"> |
|
|
</a> |
|
|
</td> |
|
|
</tr> |
|
|
</table> |
|
|
|
|
|
<table> |
|
|
<tr> |
|
|
<td> |
|
|
<a href="https://youtu.be/7rH0QP18zsk" target="_blank"> |
|
|
<img src="illustrations/videos/PEwiwzyTjX0_000589large-dapt.gif" width="100%"> |
|
|
</a> |
|
|
</td> |
|
|
<td> |
|
|
<a href="https://youtu.be/5ZNYwDGmOZI" target="_blank"> |
|
|
<img src="illustrations/videos/RASKiMoxhOE_000246_large-dapt.gif" width="100%"> |
|
|
</a> |
|
|
</td> |
|
|
</tr> |
|
|
</table> |
|
|
|
|
|
<table> |
|
|
<tr> |
|
|
<td> |
|
|
<a href="https://youtu.be/X7Ij1sc4yCE" target="_blank"> |
|
|
<img src="illustrations/videos/T7TkJVmGyts_001011_large-dapt.gif" width="100%"> |
|
|
</a> |
|
|
</td> |
|
|
<td> |
|
|
<a href="https://youtu.be/S5m2ooY6CGc" target="_blank"> |
|
|
<img src="illustrations/videos/TNZv-NBcV5U_000066large-dapt.gif" width="100%"> |
|
|
</a> |
|
|
</td> |
|
|
</tr> |
|
|
</table> |
|
|
|
|
|
|
