The camera takes a picture.
The AI makes a guess.
You make the ground truth.
Let's teach the computer about fingernails!
Single Finger Training Data Mode
Loading interface...
Computers do not see pictures like people do.
Show ONE finger inside the box.
Keep your fingertip visible.
Point your fingertip to the LEFT: β Fingertip
Keep the full fingertip visible. The large box is only a guide.
The camera took a picture. We send a smaller copy of the guide area to an AI helper on the internet. Keep faces and personal information out of the picture.
What do YOU think?
Choose your guess before asking AI.
Picture taken! πΈ
Maximum nail free edge: 3 mm
Interesting! Why do you think the answers are different?
Do you agree with AI?
Did AI find your finger correctly?
Did AI find your nail correctly?
Now let's teach AI!
Tap and compare the picture tricks below. None of these is thermal imaging.
This is what our camera captured.
Sometimes colour is not needed to find shapes.
These lines help the computer find borders.
Computers can rearrange colour information to look for patterns.
π¬ Pictures are made from tiny squares called pixels. The computer studies these tiny pieces.
The AI suggestion may already be painted. Paint, erase, undo, or clear it to make your final answer.
Not yet! Things look bigger when they are closer to the camera. First we must teach the camera how big one millimetre is. That next experiment is called calibration.
π Next Experiment: Teach the camera about millimetres using something with a known size.
Your nail marking is the correct answer. This is called ground truth.
0 / 50
Every careful example helps the experiment.
Different examples help the computer learn better.
Each tiny square is called a pixel. The computer studies these pixels.
A person can recognise a fingernail very quickly. A computer needs examples.
New STEM Word: Training
Training means showing an AI many examples so it can learn patterns.
We tell the computer rules, such as βlook for this colour.β
We show the computer examples so it can learn patterns. AI does not think like a person.
Question: Can a computer find a fingernail using colour and shape rules?
Observation: Sometimes it worked. Sometimes light, shadows and skin colour confused it.
Conclusion: We decided to teach the computer using examples. A result that did not work still taught us something useful.
The same object can look bigger close to the camera and smaller farther away.
Calibration means using something with a known size to change pixels into real measurements. We are not measuring millimetres yet.
I took 0 nail pictures.
I taught the computer where the nail was. I learned that pictures contain pixels, computer vision can find shapes, AI needs examples, and pixels are not automatically millimetres.
A camera collects light to make an image.
A pixel is one tiny coloured square in a digital picture.
Computer Vision helps a computer understand things inside pictures.
Changing a picture can make useful shapes or edges easier to see.
Training data is a collection of examples and correct answers used to teach an AI.
Calibration provides a known reference so pixels can later be compared with real units.
Educational text is stored in a language dictionary. Reviewed Tamil translations can be added without changing the activity.
Some computer programs run on this phone. Our AI helper runs on powerful computers on the internet. The phone sends the smaller finger picture and receives the AI's answer.
This experiment sends the captured finger picture to our AI helper on the internet for analysis. Keep faces and personal information out of the picture. Only the camera-guide crop is sent. The final cropped example and nail marking can be saved for the experiment, and a parent can review or delete every sample.