V

Videomae Base Finetuned Ucf101 Subset

Developed by nateraw
Video classification model based on VideoMAE architecture, fine-tuned on a subset of UCF101 with an accuracy of 85.16%
Downloads 77
Release Time : 11/10/2022

Model Overview

This model is a fine-tuned version of the VideoMAE base architecture, specifically designed for video action recognition tasks, demonstrating excellent performance on a subset of the UCF101 dataset.

Model Features

Efficient Video Understanding
Utilizes masked autoencoder pre-training strategy to effectively learn spatiotemporal features in videos
Excellent Fine-tuning Performance
Achieves 85.16% accuracy on the UCF101 subset, validating the model's effectiveness
Lightweight Architecture
Based on the base-scale VideoMAE, balancing performance and computational efficiency

Model Capabilities

Video Action Recognition
Spatiotemporal Feature Extraction
Video Content Classification

Use Cases

Intelligent Surveillance
Abnormal Behavior Detection
Identifies specific action patterns in surveillance videos
Content Analysis
Video Content Classification
Automatically labels action categories in videos
85.16% Accuracy
Featured Recommended AI Models
AIbase
Empowering the Future, Your AI Solution Knowledge Base
Š 2025AIbase