Secure On-Device Video OOD Detection Without Backpropagation

SecDOOD is a secure cloud-device collaboration framework for efficient on-device OOD detection without requiring device-side backpropagation.

_{Click the logo to read our paper on arXiv}

SecDOOD

Goal. On-device video OOD detection with no device-side backpropagation, while preserving user privacy.

Approach. A cloud-hosted HyperNetwork generates device-specific classifier weights from an encrypted summary of device features. The device runs forward-only inference; no gradients or raw data leave the device.

Workflow.

Cloud (offline): Train base model $M_g$ and HyperNetwork $H$ on ID data.
Device (online): Extract features, score channel importance, homomorphically encrypt top $\alpha%$ channels, mask the rest, and send encrypted subset.
Cloud → Device: Apply $H$ on encrypted features to produce $\Theta_d$; return to device for inference.

Privacy & Efficiency. Raw video and full features stay on device; the cloud only sees selectively encrypted channels. Dynamic channel sampling + selective encryption sharply reduce crypto and bandwidth costs with minimal accuracy loss.

Compatibility. Plug-and-play with common OOD scores (MSP, Energy, VIM) and modalities (RGB/flow/audio). Supports near- and far-OOD.

Results. On HMDB51, UCF101, EPIC-Kitchens, HAC, and Kinetics-600, SecDOOD increases AUROC and lowers FPR@95 versus local post-processing baselines—without any on-device training.

Prepare Datasets

SecDOOD is based on five public action recognition datasets (HMDB51, UCF101, EPIC-Kitchens, HAC, and Kinetics-600).

Download HMDB51 video data from link and extract. Download HMDB51 optical flow data from link and extract. The directory structure should be modified to match:

Click for details...

HMDB51
├── video
|   ├── catch
|   |   ├── *.avi
|   ├── climb
|   |   ├── *.avi
|   |── ...


├── flow
|   ├── *_flow_x.mp4
|   ├── *_flow_y.mp4
|   ├── ...

Download UCF101 video data from link and extract. Download UCF101 optical flow data from link and extract. The directory structure should be modified to match:
Click for details...
```
UCF101
├── video
|   ├── *.avi
|   |── ...


├── flow
|   ├── *_flow_x.mp4
|   ├── *_flow_y.mp4
|   ├── ...
```

Download EPIC-Kitchens video and optical flow data by

bash utils/download_epic_script.sh

Download audio data from link.

Unzip all files and the directory structure should be modified to match:

Click for details...

EPIC-KITCHENS
├── rgb
|   ├── train
|   |   ├── D3
|   |   |   ├── P22_05.wav
|   |   |   ├── P22_05
|   |   |   |     ├── frame_0000000000.jpg
|   |   |   |     ├── ...
|   |   |   ├── P22_06
|   |   |   ├── ...
|   ├── test
|   |   ├── D3
|   |   |   ├── P22_01.wav
|   |   |   ├── P22_01
|   |   |   |     ├── frame_0000000000.jpg
|   |   |   |     ├── ...
|   |   |   ├── P22_02
|   |   |   ├── ...

├── flow
|   ├── train
|   |   ├── D3
|   |   |   ├── P22_05
|   |   |   |     ├── frame_0000000000.jpg
|   |   |   |     ├── ...
|   |   |   ├── P22_06
|   |   |   ├── ...
|   ├── test
|   |   ├── D3
|   |   |   ├── P22_01
|   |   |   |     ├── frame_0000000000.jpg
|   |   |   |     ├── ...
|   |   |   ├── P22_02
|   |   |   ├── ...

Download HAC video, audio and optical flow data from link and extract. The directory structure should be modified to match:

Click for details...

HAC
├── human
|   ├── videos
|   |   ├── ...
|   ├── flow
|   |   ├── ...
|   ├── audio
|   |   ├── ...

├── animal
|   ├── videos
|   |   ├── ...
|   ├── flow
|   |   ├── ...
|   ├── audio
|   |   ├── ...

├── cartoon
|   ├── videos
|   |   ├── ...
|   ├── flow
|   |   ├── ...
|   ├── audio
|   |   ├── ...

Download Kinetics-600 video data by
```
wget -i utils/filtered_k600_train_path.txt
```
Extract all files and get audio data from video data by
```
python utils/generate_audio_files.py
```
Download Kinetics-600 optical flow data (kinetics600_flow_mp4_part_*) from link and extract (run cat kinetics600_flow_mp4_part_* > kinetics600_flow_mp4.tar.gz and then tar -zxvf kinetics600_flow_mp4.tar.gz).

Unzip all files and the directory structure should be modified to match:
Click for details...
```
Kinetics-600
├── video
|   ├── acting in play
|   |   ├── *.mp4
|   |   ├── *.wav
|   |── ...


├── flow
|   ├── acting in play
|   |   ├── *_flow_x.mp4
|   |   ├── *_flow_y.mp4
|   ├── ...
```

Dataset Splits

The splits for Multimodal Near-OOD and Far-OOD Benchmarks are provided under HMDB-rgb-flow/splits/ for HMDB51, UCF101, HAC, and Kinetics-600, and under EPIC-rgb-flow/splits/ for EPIC-Kitchens.

Methodology

SecDOOD is a cloud–device collaborative pipeline for on-device video OOD detection with zero device-side backpropagation. A cloud-hosted HyperNetwork H is trained on in-distribution (ID) data and returns device-specific classifier parameters Θ_d; the device performs forward-only inference.

Device feature extraction. Compute intermediate features and rank channels via Shapley-style contribution estimates.
Selective encryption & upload. Encrypt the top 50% most informative channels (mask the rest) and send only this encrypted subset to the cloud.
Cloud personalization. Evaluate H on the encrypted features to produce Θ_d. Raw videos and full feature tensors never leave the device.
On-device inference. Decrypt and inject Θ_d into the local head and compute OOD scores (e.g., MSP, Energy, VIM); inference is forward-only.

Privacy & efficiency. Dynamic channel selection plus selective encryption reduces bandwidth and cryptographic cost while preserving accuracy, enabling plug-and-play deployment on resource-constrained devices.

Code

The code was tested using Python 3.10.4, torch 1.11.0+cu113 and NVIDIA GeForce RTX 3090. More dependencies are in requirement.txt.

Prepare

Download Pretrained Weights

Download SlowFast model for RGB modality link and place under the HMDB-rgb-flow/pretrained_models and EPIC-rgb-flow/pretrained_models directory
Download SlowOnly model for Flow modality link and place under the HMDB-rgb-flow/pretrained_models and EPIC-rgb-flow/pretrained_models directory
Download Audio model link, rename it as vggsound_avgpool.pth.tar and place under the HMDB-rgb-flow/pretrained_models and EPIC-rgb-flow/pretrained_models directory

Multimodal Near-OOD Benchmark

HMDB51 25/26

Click for details...

cd HMDB-rgb-flow/

Train the Near-OOD model for HMDB:

python Train.py --near_ood --dataset 'HMDB' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --start_epoch 10 --use_single_pred --use_a2d --a2d_max_hellinger --a2d_ratio 0.5 --use_npmix --max_ood_hellinger --a2d_ratio_ood 0.5 --ood_entropy_ratio 0.5 --nepochs 50 --appen '' --save_best --save_checkpoint --datapath '/path/to/HMDB51/'

You can also download our provided checkpoints (HMDB_near_ood_baseline.pt, HMDB_near_ood_a2d.pt, and HMDB_near_ood_a2d_npmix.pt) from link.

Save the evaluation files for HMDB (to save evaluation files for ASH or ReAct, you should also run following line with options --use_ash or --use_react):

python Test.py --bsz 16 --num_workers 2 --near_ood --dataset 'HMDB' --appen 'a2d_npmix_best_' --resumef '/path/to/HMDB_near_ood_a2d_npmix.pt'

Evaluation for HMDB (change --postprocessor to different score functions, for VIM you should also pass --resume_file checkpoint.pt, where checkpoint.pt is the trained checkpoint):

python eval_video_flow_near_ood.py --postprocessor msp --appen 'a2d_npmix_best_' --dataset 'HMDB' --path 'HMDB-rgb-flow/'

UCF101 50/51

Click for details...

cd HMDB-rgb-flow/

Train the Near-OOD model for UCF:

python Train.py --near_ood --dataset 'UCF' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --start_epoch 10 --use_single_pred --use_a2d --a2d_max_hellinger --a2d_ratio 0.5 --use_npmix --max_ood_hellinger --a2d_ratio_ood 0.5 --ood_entropy_ratio 0.5 --nepochs 50 --appen '' --save_best --save_checkpoint --datapath '/path/to/UCF101/'

You can also download our provided checkpoints (UCF_near_ood_baseline.pt, UCF_near_ood_a2d.pt, and UCF_near_ood_a2d_npmix.pt) from link.

Save the evaluation files for UCF (to save evaluation files for ASH or ReAct, you should also run following line with options --use_ash or --use_react):

python Test.py --bsz 16 --num_workers 2 --near_ood --dataset 'UCF' --appen 'a2d_npmix_best_' --resumef '/path/to/UCF_near_ood_a2d_npmix.pt'

Evaluation for UCF (change --postprocessor to different score functions, for VIM you should also pass --resume_file checkpoint.pt, where checkpoint.pt is the trained checkpoint):

python eval_video_flow_near_ood.py --postprocessor msp --appen 'a2d_npmix_best_' --dataset 'UCF' --path 'HMDB-rgb-flow/'

EPIC-Kitchens 4/4

Click for details...

cd EPIC-rgb-flow/

Train the Near-OOD baseline model for EPIC:

python Train_Epic.py --dataset 'EPIC' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --nepochs 20 --appen '' --save_best --save_checkpoint --datapath '/path/to/EPIC-Kitchens/'

Train the Near-OOD model using A2D for EPIC:

python Train_Epic.py --dataset 'EPIC' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --start_epoch 10 --use_single_pred --use_a2d --a2d_max_hellinger --a2d_ratio 1.0 --nepochs 50 --appen '' --save_best --save_checkpoint --datapath '/path/to/EPIC-Kitchens/'

Train the Near-OOD model using A2D and NP-Mix for EPIC:

python Train_Epic.py --dataset 'EPIC' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --start_epoch 10 --use_single_pred --use_a2d --a2d_max_hellinger --a2d_ratio 0.1 --use_npmix --max_ood_hellinger --a2d_ratio_ood 0.1 --ood_entropy_ratio 0.1 --nepochs 20 --appen '' --save_best --save_checkpoint --datapath '/path/to/EPIC-Kitchens/'

You can also download our provided checkpoints (EPIC_near_ood_baseline.pt, EPIC_near_ood_a2d.pt, and EPIC_near_ood_a2d_npmix.pt) from link.

Save the evaluation files for EPIC (to save evaluation files for ASH or ReAct, you should also run following line with options --use_ash or --use_react):

python Test_near.py --bsz 16 --num_workers 2  --ood_dataset 'EPIC' --appen 'a2d_npmix_best_' --resumef '/path/to/EPIC_near_ood_a2d_npmix.pt'

Evaluation for EPIC (change --postprocessor to different score functions, for VIM you should also pass --resume_file checkpoint.pt, where checkpoint.pt is the trained checkpoint):

python eval_video_flow_near_ood.py --postprocessor msp --appen 'a2d_npmix_best_' --dataset 'EPIC' --path 'EPIC-rgb-flow/'

Kinetics-600 129/100

Click for details...

cd HMDB-rgb-flow/

Train the Near-OOD baseline model for Kinetics:

python Train.py --near_ood --dataset 'Kinetics' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --nepochs 20 --appen '' --save_best --save_checkpoint --datapath '/path/to/Kinetics-600/'

Train the Near-OOD model using A2D for Kinetics:

python Train.py --near_ood --dataset 'Kinetics' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --start_epoch 10 --use_single_pred --use_a2d --a2d_max_hellinger --a2d_ratio 1.0 --nepochs 20 --appen '' --save_best --save_checkpoint --datapath '/path/to/Kinetics-600/'

Train the Near-OOD model using A2D and NP-Mix for Kinetics:

python Train.py --near_ood --dataset 'Kinetics' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --start_epoch 10 --use_single_pred --use_a2d --a2d_max_hellinger --a2d_ratio 0.1 --use_npmix --max_ood_hellinger --a2d_ratio_ood 0.1 --ood_entropy_ratio 0.1 --nepochs 20 --appen '' --save_best --save_checkpoint --datapath '/path/to/Kinetics-600/'

You can also download our provided checkpoints (Kinetics_near_ood_baseline.pt, Kinetics_near_ood_a2d.pt, and Kinetics_near_ood_a2d_npmix.pt) from link.

Save the evaluation files for Kinetics (to save evaluation files for ASH or ReAct, you should also run following line with options --use_ash or --use_react):

python Test.py --bsz 16 --num_workers 2 --near_ood --dataset 'Kinetics' --appen 'a2d_npmix_best_' --resumef '/path/to/Kinetics_near_ood_a2d_npmix.pt'

Evaluation for Kinetics (change --postprocessor to different score functions, for VIM you should also pass --resume_file checkpoint.pt, where checkpoint.pt is the trained checkpoint):

python eval_video_flow_near_ood.py --postprocessor msp --appen 'a2d_npmix_best_' --dataset 'Kinetics' --path 'HMDB-rgb-flow/'

Multimodal Far-OOD Benchmark

HMDB51 as ID

Click for details...

cd HMDB-rgb-flow/

Train the Far-OOD baseline model for HMDB:

python Train.py --dataset 'HMDB' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --nepochs 50 --appen '' --save_best --save_checkpoint --datapath '/path/to/HMDB51/'

Train the Far-OOD model using A2D and NP-Mix for HMDB:

python Train.py --dataset 'HMDB' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --start_epoch 10 --use_single_pred --use_a2d --a2d_max_hellinger --a2d_ratio 0.1 --use_npmix --max_ood_hellinger --a2d_ratio_ood 0.1 --ood_entropy_ratio 0.1 --nepochs 50 --appen '' --save_best --save_checkpoint --datapath '/path/to/HMDB51/'

You can also download our provided checkpoints (HMDB_far_ood_baseline.pt and HMDB_far_ood_a2d_npmix.pt) from link.

Save the evaluation files for HMDB (to save evaluation files for ASH or ReAct, you should also run following line with options --use_ash or --use_react, same for other datasets):

python Test.py --bsz 16 --num_workers 2 --dataset 'HMDB' --appen 'a2d_npmix_best_' --resumef '/path/to/HMDB_far_ood_a2d_npmix.pt'

Save the evaluation files for UCF:

python Test.py --bsz 16 --num_workers 2 --far_ood --dataset 'HMDB' --ood_dataset 'UCF' --appen 'a2d_npmix_best_' --resumef '/path/to/HMDB_far_ood_a2d_npmix.pt'

Save the evaluation files for HAC:

python Test.py --bsz 16 --num_workers 2 --far_ood --dataset 'HMDB' --ood_dataset 'HAC' --appen 'a2d_npmix_best_' --resumef '/path/to/HMDB_far_ood_a2d_npmix.pt'

Save the evaluation files for Kinetics:

python Test.py --bsz 16 --num_workers 2 --far_ood --dataset 'HMDB' --ood_dataset 'Kinetics' --appen 'a2d_npmix_best_' --resumef '/path/to/HMDB_far_ood_a2d_npmix.pt'

Save the evaluation files for EPIC:

cd EPIC-rgb-flow/

python Test_far.py --bsz 16 --num_workers 2 --far_ood --dataset 'HMDB' --ood_dataset 'EPIC' --appen 'a2d_npmix_best_' --resumef '/path/to/HMDB_far_ood_a2d_npmix.pt'

Evaluation for UCF (change --postprocessor to different score functions, for VIM you should also pass --resume_file checkpoint.pt, where checkpoint.pt is the trained checkpoint, change --ood_dataset to UCF, EPIC, HAC, or Kinetics):

python eval_video_flow_far_ood.py --postprocessor msp --appen 'a2d_npmix_best_' --dataset 'HMDB' --ood_dataset 'UCF' --path 'HMDB-rgb-flow/'

Kinetics as ID

Click for details...

cd HMDB-rgb-flow/

Train the Far-OOD baseline model for Kinetics:

python Train.py --dataset 'Kinetics' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --nepochs 10 --appen '' --save_best --save_checkpoint --datapath '/path/to/Kinetics-600/'

Train the Far-OOD model using A2D and NP-Mix for Kinetics:

python Train.py --dataset 'Kinetics' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --start_epoch 3 --use_single_pred --use_a2d --a2d_max_hellinger --a2d_ratio 0.1 --use_npmix --max_ood_hellinger --a2d_ratio_ood 0.1 --ood_entropy_ratio 0.1 --nepochs 10 --appen '' --save_best --save_checkpoint --datapath '/path/to/Kinetics-600/'

You can also download our provided checkpoints (Kinetics_far_ood_baseline.pt and Kinetics_far_ood_a2d_npmix.pt) from link.

Save the evaluation files for Kinetics (to save evaluation files for ASH or ReAct, you should also run following line with options --use_ash or --use_react, same for other datasets):

python Test.py --bsz 16 --num_workers 2 --dataset 'Kinetics' --appen 'a2d_npmix_best_' --resumef '/path/to/Kinetics_far_ood_a2d_npmix.pt'

Save the evaluation files for HMDB:

python Test.py --bsz 16 --num_workers 2 --far_ood --dataset 'Kinetics' --ood_dataset 'HMDB' --appen 'a2d_npmix_best_' --resumef '/path/to/Kinetics_far_ood_a2d_npmix.pt'

Save the evaluation files for UCF:

python Test.py --bsz 16 --num_workers 2 --far_ood --dataset 'Kinetics' --ood_dataset 'UCF' --appen 'a2d_npmix_best_' --resumef '/path/to/Kinetics_far_ood_a2d_npmix.pt'

Save the evaluation files for HAC:

python Test.py --bsz 16 --num_workers 2 --far_ood --dataset 'Kinetics' --ood_dataset 'HAC' --appen 'a2d_npmix_best_' --resumef '/path/to/cKinetics_far_ood_a2d_npmix.pt'

Save the evaluation files for EPIC:

cd EPIC-rgb-flow/

python Test_far.py --bsz 16 --num_workers 2 --far_ood --dataset 'Kinetics' --ood_dataset 'EPIC' --appen 'a2d_npmix_best_' --resumef '/path/to/Kinetics_far_ood_a2d_npmix.pt'

Evaluation for UCF (change --postprocessor to different score functions, for VIM you should also pass --resume_file checkpoint.pt, where checkpoint.pt is the trained checkpoint, change --ood_dataset to UCF, EPIC, HAC, or HMDB):

python eval_video_flow_far_ood.py --postprocessor msp --appen 'a2d_npmix_best_' --dataset 'Kinetics' --ood_dataset 'UCF' --path 'HMDB-rgb-flow/'

Multimodal Near-OOD Benchmark with Video, Audio, and Optical Flow

EPIC-Kitchens 4/4

Click for details...

cd EPIC-rgb-flow/

Train the Near-OOD baseline model for EPIC:

python Train_audio_epic.py --dataset 'EPIC' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --nepochs 20 --appen '' --save_best --save_checkpoint --datapath '/path/to/EPIC-Kitchens/'

Train the Near-OOD model using A2D and NP-Mix for EPIC:

python Train_audio_epic.py --dataset 'EPIC' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --start_epoch 10 --use_single_pred --use_a2d --a2d_max_hellinger --a2d_ratio 0.5 --use_npmix --max_ood_hellinger --a2d_ratio_ood 0.5 --ood_entropy_ratio 0.5 --nepochs 20 --appen '' --save_best --save_checkpoint --datapath '/path/to/EPIC-Kitchens/'

You can also download our provided checkpoints (EPIC_near_ood_vfa_baseline.pt and EPIC_near_ood_vfa_a2d_npmix.pt) from link.

Save the evaluation files for EPIC (to save evaluation files for ASH or ReAct, you should also run following line with options --use_ash or --use_react):

python Test_audio_epic.py --bsz 16 --num_workers 2  --ood_dataset 'EPIC' --appen 'a2d_npmix_best_' --resumef '/path/to/EPIC_near_ood_vfa_a2d_npmix.pt'

Evaluation for EPIC (change --postprocessor to different score functions, for VIM you should also pass --resume_file checkpoint.pt, where checkpoint.pt is the trained checkpoint):

python eval_video_flow_near_ood.py --postprocessor msp --appen 'vfa_a2d_npmix_best_' --dataset 'EPIC' --path 'EPIC-rgb-flow/'

Kinetics-600 129/100

Click for details...

cd HMDB-rgb-flow/

Train the Near-OOD baseline model for Kinetics:

python Train_audio.py --near_ood --dataset 'Kinetics' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --nepochs 20 --appen '' --save_best --save_checkpoint --datapath '/path/to/Kinetics-600/'

Train the Near-OOD model using A2D and NP-Mix for Kinetics:

python Train_audio.py --near_ood --dataset 'Kinetics' --lr 0.0001 --seed 0 --bsz 16 --num_workers 10 --start_epoch 10 --use_single_pred --use_a2d --a2d_max_hellinger --a2d_ratio 0.5 --use_npmix --max_ood_hellinger --a2d_ratio_ood 0.5 --ood_entropy_ratio 0.5 --nepochs 20 --appen '' --save_best --save_checkpoint --datapath '/path/to/Kinetics-600/'

You can also download our provided checkpoints (Kinetics_near_ood_vfa_baseline.pt and Kinetics_near_ood_vfa_a2d_npmix.pt) from link.

Save the evaluation files for Kinetics (to save evaluation files for ASH or ReAct, you should also run following line with options --use_ash or --use_react):

python Test_audio.py --bsz 16 --num_workers 2 --near_ood --dataset 'Kinetics' --appen 'a2d_npmix_best_' --resumef '/path/to/Kinetics_near_ood_a2d_npmix.pt'

Evaluation for Kinetics (change --postprocessor to different score functions, for VIM you should also pass --resume_file checkpoint.pt, where checkpoint.pt is the trained checkpoint):

python eval_video_flow_near_ood.py --postprocessor msp --appen 'vfa_a2d_npmix_best_' --dataset 'Kinetics' --path 'HMDB-rgb-flow/'

Contact

If you have any questions, please send an email to [email protected] or make an issue :)

Citation

@article{li2025secure,
  title={Secure on-device video ood detection without backpropagation},
  author={Li, Shawn and Cai, Peilin and Zhou, Yuxiao and Ni, Zhiyu and Liang, Renjie and Qin, You and Nian, Yi and Tu, Zhengzhong and Hu, Xiyang and Zhao, Yue},
  journal={arXiv preprint arXiv:2503.06166},
  year={2025}
}

Name		Name	Last commit message	Last commit date
Latest commit History 11 Commits
EPIC-rgb-flow		EPIC-rgb-flow
HMDB-rgb-flow		HMDB-rgb-flow
imgs		imgs
static		static
utils		utils
.gitignore		.gitignore
README.md		README.md
README_.md		README_.md
README_template.md		README_template.md
eval_video_flow_far_ood.py		eval_video_flow_far_ood.py
eval_video_flow_near_ood.py		eval_video_flow_near_ood.py
index.html		index.html
metrics.py		metrics.py
requirement.txt		requirement.txt
test_eval.sh		test_eval.sh
train.sh		train.sh

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Repository files navigation

Secure On-Device Video OOD Detection Without Backpropagation

SecDOOD

Prepare Datasets

Dataset Splits

Methodology

Code

Prepare

Download Pretrained Weights

Multimodal Near-OOD Benchmark

HMDB51 25/26

UCF101 50/51

EPIC-Kitchens 4/4

Kinetics-600 129/100

Multimodal Far-OOD Benchmark

HMDB51 as ID

Kinetics as ID

Multimodal Near-OOD Benchmark with Video, Audio, and Optical Flow

EPIC-Kitchens 4/4

Kinetics-600 129/100

Contact

Citation

About

Uh oh!

Releases

Packages

Languages

Dystopians/SecDOOD

Folders and files

Latest commit

History

Repository files navigation

Secure On-Device Video OOD Detection Without Backpropagation

SecDOOD

Prepare Datasets

Dataset Splits

Methodology

Code

Prepare

Download Pretrained Weights

Multimodal Near-OOD Benchmark

HMDB51 25/26

UCF101 50/51

EPIC-Kitchens 4/4

Kinetics-600 129/100

Multimodal Far-OOD Benchmark

HMDB51 as ID

Kinetics as ID

Multimodal Near-OOD Benchmark with Video, Audio, and Optical Flow

EPIC-Kitchens 4/4

Kinetics-600 129/100

Contact

Citation

About

Resources

Uh oh!

Stars

Watchers

Forks

Releases

Packages 0

Languages

Packages