Facial Recognition: How TensorFlow Knows It’s You

You upload a photo to Facebook. It suggests tagging your friend. How? It uses a Convolutional Neural Network (CNN).

Deep Learning Layers

Imagine a stack of translucent filters.

1. Layer 1: Finds Edges. (Lines, curves).
2. Layer 2: Finds Shapes. (Circles, triangles).
3. Layer 3: Finds Features. (Eyes, Noses).
4. Layer 4: Finds Faces. (Steve, Sarah).

The computer doesn’t know what a “Nose” is. It just knows that “Pattern #453” usually appears in the middle of “Pattern #890” (Face).

Training Data

To teach the AI, you show it 1 million photos.

• “This is Steve.”
• “This is NOT Steve.” Over time, it adjusts its internal math weights until it gets it right.

Robot Object Detection

We use TensorFlow in FTC.

• We show the robot 1,000 photos of a Yellow Foam Ring.
• We train the model on our laptop.
• We upload the .tflite file to the robot.
• Now, during the match, the robot drives around looking for “Pattern Yellow-Ring.” It’s the exact same technology used by the FBI, just applied to foam toys.