Skip to main content

Documentation Index

Fetch the complete documentation index at: https://mintlify.com/ilirosmanaj/detect_kermit/llms.txt

Use this file to discover all available pages before exploring further.

Detect Kermit is a Python machine learning project that trains a binary image classifier to identify whether Kermit the Frog from The Muppets TV show appears in a given image or video frame. It uses ImageAI’s CustomImagePrediction (ResNet architecture) on top of TensorFlow, and supports inference on both static images and full video files.

Introduction

Understand how the project works, its architecture, and what you can do with it.

Quickstart

Install dependencies, load the pre-trained model, and run your first prediction in minutes.

Training Guide

Train or retrain the ResNet model on your own dataset of Kermit and non-Kermit images.

Prediction Guide

Run predictions on images and video files using the command-line interface.

How It Works

1

Prepare your dataset

Collect and organise training images into kermit and no-kermit folders. Use the helper scripts to extract frames from video, rotate images for augmentation, and download extra images from Google.
2

Train the model

Run imageai_build_model.py to train a ResNet model using ImageAI’s CustomImagePrediction. The trained weights are saved as kermit_finder.h5.
3

Run predictions

Use kermit_model_evaluation.py with the -t image or -t video flag to classify a single image or annotate every second of a video with the model’s prediction.
A pre-trained model (data/images/models/kermit_finder.h5) and class map (data/images/json/model_class.json) are included in the repository via Git LFS, so you can run predictions without retraining.

Build docs developers (and LLMs) love