This repo is a fork of the original. It contains converted TensorFlow JS models.

This project uses a MobileNet architecture for classifying grayscale images. The model has been optimized to handle greyscale images efficiently, leveraging the depth and speed of MobileNet, which is typically used for lightweight image classification tasks. The project includes a data pipeline, model training, evaluation, and visualization of results, including a accuracy_loss graph to assess model performance.

Loads the model and predict on output.ndjson (an airplane sketch).

python3 load_model.py

Graph model. Converted to directory tfjs. Layered model conversion is not supported by the conversion tool.

Include local tfjs or from cdn:

<script src="./js/tfjs@4.22.0.js"></script>

Then load it:

async loadModel() { tf.loadGraphModel("./models/model.json").then(classifier => { // ... }); }