OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

3D Resnets Pytorch

3D Resnets Pytorch is a Python and PyTorch research codebase for training, testing and fine-tuning 3D ResNet and R(2+1)D models for action recognition in video. It includes pretrained weights associated with Kinetics-700, Moments in Time and STAIR-Actions, including merged-dataset models.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

3D Resnets Pytorch — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you need research code and pretrained 3D convolutional models for video action recognition, or a documented training and fine-tuning workflow for the named datasets.

Who it is for

  • Researchers comparing spatiotemporal video-recognition models.
  • Developers building video action-classification workflows who can prepare datasets and work with Python and PyTorch.

Documented uses

  • Classify actions in video clips using 3D convolutional models.
  • Fine-tune pretrained models on a target dataset.
  • Train and test models on supported datasets, including Kinetics, Moments in Time, ActivityNet, UCF-101 and HMDB-51.
  • Prepare video frames and annotations for a dataset workflow.
Before you choose

Requirements & limits

The repository is public and identifies its license as MIT. The README lists Python 3, PyTorch 0.4 or later, FFmpeg and FFprobe as requirements.

  • The project is code and model weights, not a hosted video-analysis application; using it requires a Python and PyTorch environment.
  • Dataset preparation is part of the workflow: the README describes extracting JPG frames and creating annotations, with dataset-specific inputs.
  • Pretrained checkpoints vary by model depth and training dataset; the README gives class-count options for fine-tuning.
  • The README directs users seeking to reproduce the CVPR 2018 paper to the CVPR2018 branch.
  • The updated scripts describe distributed-training support and R(2+1)D models.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Prepare UCF-101 frames and annotations

Convert downloaded UCF-101 videos to JPG frames and create the dataset annotation file used by the repository.

Before you start

  • Python 3, PyTorch 0.4 or later, FFmpeg and FFprobe.
  • UCF-101 videos and train/test splits.
  • The repository's util_scripts and the UCF-101 annotation files, including classInd.txt and trainlist/testlist split files.

Steps

  1. Download the UCF-101 videos and train/test splits from the dataset source named in the README.

  2. Convert the AVI videos into JPG frames with util_scripts/generate_video_jpgs.py using the ucf101 dataset argument.

  3. Generate an annotation JSON file with util_scripts/ucf101_json.py, providing the annotation directory, JPG-frame directory and destination JSON path.

Expected outcome

A JPG-frame directory and a dataset annotation JSON file prepared for the repository's UCF-101 workflow.

Guide limitations

  • This recipe covers data preparation only; it does not document model training or inference steps.
  • The UCF-101 videos, splits and annotations must be obtained separately.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
16 October 2026
Original website title
kenshohara/3D-ResNets-PyTorch · Official homepage title · retained in its source language
How we review sourcesSuggest a correction