JUHE API Marketplace
Back to datasets
Dataset assetOpen Source CommunityVideo AnalysisAction Recognition

UCF-101, HMDB-51

UCF‑101 and HMDB‑51 are two video datasets used for training and testing video‑processing models. UCF‑101 contains 101 action categories with over 100 videos per category. HMDB‑51 includes 51 action categories with at least 101 videos per category.

Source
github
Created
May 3, 2020
Updated
May 8, 2024
Signals
361 views
Availability
Linked source ready
Overview

Dataset description and usage context

Dataset Overview

Dataset Names

  • UCF‑101
  • HMDB‑51

Dataset Download

Dataset Structure

UCF‑101
UCF‑101
├── ApplyEyeMakeup
│   ├── v_ApplyEyeMakeup_g01_c01.avi
│   └── ...
├── ApplyLipstick
│   ├── v_ApplyLipstick_g01_c01.avi
│   └── ...
├── Archery
│   ├── v_Archery_g01_c01.avi
│   └── ...
HMDB‑51
HMDB51
├── brush_hair
│   ├── April_09_brush_hair_u_nm_np1_ba_goo_0.avi
│   └── ...
├── cartwheel
│   ├── (Rad)Schlag_die_Bank!_cartwheel_f_cm_np1_le_med_0.avi
│   └── ...
├── catch
│   ├── 96-_Torwarttraining_1_catch_f_cm_np1_le_bad_0.avi
│   └── ...

Pre‑processing Steps

UCF‑101
  1. Convert video files to JPG frames using utils/video2jpg_ucf101_hmdb51.py.
  2. Generate frame‑count files with utils/n_frames_ucf101_hmdb51.py.
HMDB‑51
  1. Convert video files to JPG frames using utils/video2jpg_ucf101_hmdb51.py.
  2. Generate frame‑count files with utils/n_frames_ucf101_hmdb51.py.
  3. Generate annotation files using utils/hmdb_gen_txt.py.

Post‑processing Data Structure

UCF‑101
UCF101_n_frames
├── ApplyEyeMakeup
│   ├── v_ApplyEyeMakeup_g01_c01
│   │   ├── image_00001.jpg
│   │   └── ...
│   │   └── n_frames
│   └── ...
├── ApplyLipstick
│   └── ...
└── ...
HMDB‑51
hmdb51_n_frames
├── brush_hair
│   ├── April_09_brush_hair_u_nm_np1_ba_goo_0
│   │   ├── image_00001.jpg
│   │   └── ...
│   │   └── n_frames
│   └── ...
├── cartwheel
│   └── ...
└── ...

Data Loading

PyTorch loading examples are provided using the HMDBDataset class.

Citation

Dataset processing code is adapted from 3D‑ResNets‑PyTorch.

Need downstream help?

Pair the dataset with AI analysis and content workflows.

Once the source passes your review, move straight into summarization, transformation, report drafting, or presentation generation with the JuheAI toolchain.

Explore AI studio