What license covers this dataset?

The code and data structure are under MIT. The media assets (GIFs and thumbnails) come from Gym visual and carry separate attribution terms described in NOTICE.md, so check that file before redistributing images.

Do I need a server to browse the exercises?

No. The `index.html` file is a fully client-side browser with search, filtering, and infinite scroll that runs by just opening it in a modern browser.

How do I get this data into my own database?

Open `setup.html`, pick your database (SQL Server, PostgreSQL, MySQL, or SQLite), and it generates a ready-to-run `.sql` file with all 1,324 INSERT statements in the browser.

Which languages are the exercise instructions available in?

English, Spanish, Italian, Turkish, Russian, Chinese, Hindi, Polish, Korean, and French, both as full text strings and as split step arrays per exercise.

Does the dataset include video, not just GIFs?

No, each exercise ships only an animation GIF and a 180×180 static thumbnail image, not full video files.

Is there a schema to validate the JSON data?

Yes, `data/exercises.schema.json` is a JSON Schema (Draft 2020-12) describing every field and type, usable with any standard JSON Schema validator.

Exercises Dataset: A 1,324-exercise dataset

This repository packages 1,324 fitness exercises into a ready-to-use dataset: each entry includes an animation GIF, a 180×180 thumbnail, category and body-part tags, target and secondary muscle data, equipment requirements, and step-by-step instructions translated into 10 languages. It's built for developers building fitness or workout apps, researchers working on exercise recognition or recommendation systems, and anyone who needs a structured exercise database without scraping or licensing one from scratch. The dataset already powers LogPress, an AI-assisted workout tracker, and is designed to drop into any backend.

Key features

Full exercise coverage: 1,324 exercises spanning upper arms, upper legs, back, waist, chest, shoulders, lower legs, lower arms, cardio, and neck.
Media included: an animation GIF and a 180×180 thumbnail for every exercise, sourced from Gym visual with attribution fields baked into each record.
Multilingual instructions: step-by-step guidance in English, Spanish, Italian, Turkish, Russian, Chinese, Hindi, Polish, Korean, and French, provided both as full text and as split instruction arrays.
Structured metadata: category, target muscle, muscle group, secondary muscles, and equipment type (barbell, dumbbell, cable, body weight, and more) for filtering and recommendation logic.
JSON Schema validation: a Draft 2020-12 schema file describes every field, so you can validate the dataset or any additions you make.
No-server browser: index.html is a client-side exercise explorer with live search, filters, and infinite scroll, useful for previewing the data before integrating it.
Developer setup guide: setup.html generates CREATE TABLE SQL and INSERT statements for SQL Server, PostgreSQL, MySQL, and SQLite, plus copy-paste API client code in JavaScript, Python, C#, Java, PHP, Go, and cURL.
Equipment-free options: roughly a quarter of the exercises require no equipment, useful for at-home workout features.

Ideal use cases

This dataset fits well if you're building a workout tracker, gym app, or fitness content platform and need pre-structured exercise data instead of manually entering hundreds of records. It also works for machine learning projects around exercise classification, recommendation engines based on equipment or muscle group, or research into fitness content across languages. The multilingual instructions make it a reasonable starting point for apps targeting non-English-speaking users.

It's not a fit if you need video (only GIFs and static thumbnails are provided), live coaching or form-checking features, or exercises outside the 1,324 covered (there's no mechanism here for user-submitted or dynamically expanding content). If you need commercial redistribution rights to the media assets beyond what's described in the license and NOTICE files, you'll need to check terms with Gym visual directly, since the media has its own attribution requirements separate from the MIT-licensed code and data structure.

Installation

There's no package to install; you use the repository's files directly.

Clone the repo:

git clone https://github.com/hasaneyldrm/exercises-dataset.git
cd exercises-dataset

Open the interactive browser locally (no server needed):

open index.html

Or open setup.html in a browser to generate database import scripts and API client snippets. It walks through:

Choosing a database (SQL Server, PostgreSQL, MySQL, or SQLite) to generate a .sql file with all 1,324 INSERT statements.
Copying ready-made API client code (JavaScript, Python, C#, Java, PHP, Go, cURL) after entering your backend base URL.
Using a structured LLM prompt (pick your framework and database) to generate a REST API around the dataset, with support for Express.js, FastAPI, ASP.NET Core, Spring Boot, Laravel, or Gin.

To use the raw data directly, read data/exercises.json (a JSON array of exercise records) and validate it against data/exercises.schema.json with any standard JSON Schema validator. Image and GIF paths in each record point to files in images/ and videos/.