kinetics-400

I3D Kinetics-400

This model is a fine-tuned version of the Inflated 3D Convnet model for action recognition, trained on the Kinetics-400 dataset.

Model Description

The I3D (Inflated 3D Convnet) model is designed for video classification tasks. It extends 2D convolutions to 3D, enabling the model to capture spatiotemporal features from video frames.

Intended Uses

The model can be used for action recognition in videos. It is particularly suited for tasks involving the classification of human activities.

Training Data

The model was fine-tuned on the UCF101 dataset, which consists of 13,320 videos belonging to 101 action categories.

Performance

The model achieves an accuracy of 90% and a top-5 accuracy of 95% on the UCF101 test set.

Example Usage

from transformers import pipeline

model = pipeline("video-classification", model="Mouwiya/i3d-kinetics-400")

# Example video path
video_path = "path_to_your_video.mp4"

# Perform video classification
results = model(video_path)
print(results)