Train a model.
Get a XIAO ML Kit.
Build a computer vision model, deploy it to the edge, and we'll ship you a Seeed Studio XIAO ESP32S3 Sense kit for free.
What is Hack-Vision?
Hack-Vision is a You Ship, We Ship (YSWS) program by Hack Club. We want to see what you can build with Computer Vision.
Whether it's a model that detects if your cat is on your desk, a gesture-controlled game, or a smart doorbell—if you train a custom CV model and build a project around it, we'll reward you with hardware to take it to the next level.
- Train your own model or you can also fine-tune a pre-trained model
- Write the code to run it
- Ship it and get a free hardware kit
See the unseen.
How to get your kit
Follow these steps to ship your project and claim your hardware.
1. Collect & Trainrain
Gather a dataset and train a custom computer vision model. You can use tools like Teachable Machine, Roboflow, or write your own PyTorch/TensorFlow code.
2. Build an App
Write a script or web app that uses your trained model to do something cool. It needs to be a functional project, not just a Jupyter notebook.
3. Ship It
Open source your code on GitHub, record a demo video showing it working, and submit it to the Hack Club team.
Crash Course: ML 101
Never trained a model before? Don't panic. Here is the hacker's guide to teaching rocks how to think.
1. Data is Everything
A model is only as good as its data. Gather hundreds of images of what you want to detect. If you're building an object detector, you'll need to draw bounding boxes around the objects in every image (tools like Roboflow make this easy).
2. Pick Your Weapon
Beginner? Use Google's Teachable Machine for zero-code classification. Intermediate? Try Ultralytics YOLOv8 or Roboflow for object detection. Hardcore? Write custom PyTorch or TensorFlow scripts.
3. Train & Iteraterate
Feed your images into the neural network. The model will guess, calculate its error (loss), and adjust its weights. Watch the loss curve go down. If it's not accurate enough, you usually need more diverse data.
4. Export to the Edge
Once trained, export your model into a lightweight format like TensorFlow Lite (.tflite) or ONNX. These optimized formats are small enough to run directly on microcontrollers like the XIAO ESP32S3.
Frequently Asked Questions
Who can participate?
Any high schooler (or younger) anywhere in the world! You must be a teenager(18 or below) to receive the hardware.
Can I use OpenAI/Gemini/Claude APIs?
No! The goal of this YSWS is to learn how to train your own models. Using a pre-trained API defeats the purpose. You must train a custom model (e.g., using Teachable Machine, YOLO, PyTorch, etc.).
Do I need to pay for shipping?
Nope! Hack Club covers the cost of the hardware and international shipping. It's 100% free.
What if I've never done ML before?
That's exactly why we made this! We recommend starting with Google's Teachable Machine—it lets you train a model in your browser with no code, then export it to use in a project.