ML, Web

PoseNet Camera Feed Demo

Portfolio-title

PoseNet Camera Feed Demo

With PoseNet running on TensorFlow.js anyone with a webcam-equipped desktop or phone can experience the technology right from within a web browser.

Since PoseNet on TensorFlow.js runs in the browser, no pose data ever leaves a user’s computer.

Check it out
ML

LSTM Nietzsche Text Generation

Portfolio-title

LSTM Nietzsche Text Generation

Using RNNs and LSTMs that were trained on a corpus of writings by Nietzsche to sample and generate completely new sequences of text.

See the code
Other

TKS + Walmart Recommendation

Portfolio-title
ML, Web

Skin Lesion Classifier

Portfolio-title

Skin Lesion Classifier

A classifier written in native Keras and converted to TensorFlow.js that can be used to classify seven different types of skin lesions.

This model uses the MobileNet architecture and was retrained on the HAM10000 dataset which contains approximately 10,000 documented images.

The model obtains an accuracy of approximately 80%, with a top-3 accuracy of approximately 94%.

Check it out Read the article
ML

MNIST Convolutional Neural Network

Portfolio-title

MNIST Convolutional Neural Network

How CNNs work, real-world applications, and how to code CNNs in Python.

The MNIST dataset is a collection of 70,000 28x28 images of handwritten digits.

This model achieves an accuracy of approximately 98 to 99 percent.

Read the article See the code