VideoMAE - Base - Finetuned UCF - Crime Binary Balanced v6 Open - Source Video Analysis Model

Home

Videomae Base Finetuned Ucf Crimevbinary Balancedv6

Developed by shahadalll

A video analysis model fine-tuned based on MCG-NJU/videomae-base, specializing in anomaly behavior detection tasks

Video Processing

Transformers

#Video Anomaly Detection #Crime Recognition #High-Precision Classification

Downloads 133

Release Time : 1/8/2025

Model Overview

This model is a video understanding model based on the VideoMAE architecture, fine-tuned for video anomaly detection tasks. It demonstrates high accuracy (84.75%) and AUC value (0.9263) on the evaluation set.

Model Features

High-Precision Anomaly Detection

Achieves 84.75% accuracy and 0.9263 AUC on the evaluation set, demonstrating excellent anomaly detection capabilities

Based on VideoMAE Architecture

Utilizes an efficient video masked autoencoder pre-training method, effectively learning spatiotemporal features in videos

Balanced Training

The 'balancedv6' in the model name indicates the use of balanced training strategies, likely optimized for class imbalance issues

Model Capabilities

Video Content Analysis

Anomaly Behavior Detection

Spatiotemporal Feature Extraction

Use Cases

Public Safety

Surveillance Video Analysis

Automatically detects anomalies or suspicious behaviors in surveillance videos

Can identify 84.75% of anomalous events

Smart Retail

Store Anomaly Detection

Detects theft, violence, and other anomalous behaviors in stores

🚀 videomae-base-finetuned-ucf-crimevbinary-balancedv6

This model is a fine - tuned version of [MCG - NJU/videomae - base](https://huggingface.co/MCG - NJU/videomae - base), which can achieve good performance on the evaluation set.

🚀 Quick Start

This model is a fine-tuned version of MCG-NJU/videomae-base on an unknown dataset. It achieves the following results on the evaluation set:

Loss: 0.6175
Accuracy: 0.8475
Precision: 0.8572
Recall: 0.8475
Auc: 0.9263

✨ Features

Library Name: transformers
License: cc - by - nc - 4.0
Base Model: MCG - NJU/videomae - base
Tags: generated_from_trainer
Metrics: accuracy, precision, recall

Property	Details
Model Type	videomae - base - finetuned - ucf - crimevbinary - balancedv6
Training Data	Unknown

📚 Documentation

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 3e - 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_steps: 500
num_epochs: 30

Training results

Training Loss	Epoch	Step	Validation Loss	Accuracy	Precision	Recall	Auc
0.5676	1.0	165	0.5685	0.6608	0.6975	0.6608	0.8059
0.5645	2.0	330	0.3607	0.8481	0.8509	0.6608	0.9362
0.5645	2.0	330	0.3607	0.8481	0.8509	0.8481	0.9362
0.4638	3.0	495	0.5746	0.8021	0.8226	0.8021	0.8919
0.5765	4.0	660	0.3634	0.8622	0.8626	0.8622	0.9262
0.4146	5.0	825	0.5092	0.8092	0.8302	0.8092	0.9225
0.4319	6.0	990	0.4897	0.8799	0.8811	0.8799	0.9316
0.3396	7.0	1155	0.5233	0.8587	0.8599	0.8587	0.9257
0.1152	8.0	1320	0.6568	0.8763	0.8767	0.8763	0.9190
0.0578	9.0	1485	0.6344	0.8693	0.8716	0.8693	0.9321
0.0029	10.0	1650	0.7321	0.8728	0.8739	0.8728	0.9280
0.1363	11.0	1815	0.7399	0.8622	0.8636	0.8622	0.9215

Framework versions

Transformers 4.47.0.dev0
Pytorch 2.5.1+cu124
Datasets 3.1.0
Tokenizers 0.20.3

📄 License

This model is released 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