Sabitlenmiş Tweet
Built something fun while working on my hand sign recognition project.
I can now perform the Shadow Clone hand sign and have the system trigger the jutsu in real time.
The pipeline is fairly simple in concept:
- The webcam tracks the particular hand landmarks using MediaPipe
- Those landmarks are converted into feature vectors
- A trained model predicts the hand sign
- The UI then maps that prediction to a visual effect animation mimicking the clone effect
What looks simple in the demo took quite a bit of iteration.
Handling false positives, stabilizing predictions, and making the effect feel responsive were the main challenges.
Right now this is working for a single jutsu.
#machinelearning #computervision #mediapipe #opencv #buildinpublic
English
