videomae-base-finetuned-kinetics-0408_final_45sec_org Open-source Video Understanding Model

Videomae Base Finetuned Kinetics 0408 Final 45sec Org

Developed by d2o2ji

A video understanding model fine-tuned based on MCG-NJU/videomae-base-finetuned-kinetics, achieving an accuracy of 90.97% on the evaluation set

Video Processing

Transformers

#Video Action Recognition #High Accuracy #Fine-tuned Model

Downloads 26

Release Time : 4/7/2025

Model Overview

This model is a video classification model based on the VideoMAE architecture, specifically optimized for action recognition tasks.

Model Features

High Accuracy

Achieves 90.97% accuracy on the evaluation set

Based on VideoMAE Architecture

Utilizes the Video Masked Autoencoder pre-training framework

Long Video Processing

Supports video inputs up to 45 seconds in length (inferred from the model name)

Model Capabilities

Video Action Recognition

Temporal Behavior Analysis

Video Content Understanding

Use Cases

Video Analysis

Action Recognition

Recognize human actions in videos

90.97% accuracy

Behavior Analysis

Analyze temporal behavior patterns in videos

Intelligent Surveillance

Abnormal Behavior Detection

Detect unusual actions in surveillance videos

🚀 videomae-base-finetuned-kinetics-0408_final_45sec_org

This model is a fine - tuned version of [MCG - NJU/videomae - base - finetuned - kinetics](https://huggingface.co/MCG - NJU/videomae - base - finetuned - kinetics) on an unknown dataset. It can achieve high accuracy in video - related tasks, providing reliable performance for video analysis.

🚀 Quick Start

This model is a fine - tuned version of [MCG - NJU/videomae - base - finetuned - kinetics](https://huggingface.co/MCG - NJU/videomae - base - finetuned - kinetics) on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.3145
Accuracy: 0.9097

📚 Documentation

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

🔧 Technical Details

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e - 05
train_batch_size: 8
eval_batch_size: 8
seed: 42
optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon = 1e - 08 and optimizer_args = No additional optimizer arguments
lr_scheduler_type: linear
lr_scheduler_warmup_ratio: 0.05
training_steps: 63500

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy
0.1938	0.0100	636	0.2249	0.9151
0.0531	1.0100	1272	0.3014	0.9234
0.0298	2.0100	1908	0.1246	0.9594
1.4805	3.0100	2544	0.1870	0.9410
0.3009	4.0100	3180	0.1788	0.9280
0.0252	5.0100	3816	0.2540	0.9244
0.0565	6.0100	4452	0.3528	0.9170
0.2155	7.0100	5088	0.4423	0.8902
0.0021	8.0100	5724	0.2602	0.9373
0.0006	9.0100	6360	0.3773	0.9225
0.0025	10.0100	6996	0.1920	0.9465
0.0031	11.0100	7632	0.1979	0.9530
0.0003	12.0100	8268	0.2409	0.9410
0.1511	13.0100	8904	0.7389	0.8690
0.0002	14.0100	9540	0.2832	0.9345
0.0014	15.0100	10176	0.2735	0.9410
0.0033	16.0100	10812	0.2601	0.9400
0.0039	17.0100	11448	0.2795	0.9428

Framework versions

Transformers 4.48.1
Pytorch 2.5.1+cu121
Datasets 3.2.0
Tokenizers 0.21.0

📄 License

This model is licensed under the cc - by - nc - 4.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご