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 relies on a specific directory layout and two artefact files (a .h5 model checkpoint and a .json class map) to function correctly. This page documents all configuration surfaces: the expected folder tree, the class-mapping JSON, the trained model file, pinned Python dependencies, and the EXECUTION_PATH convention used throughout the codebase.

​
Directory structure

The repository must be laid out as follows before running training or inference. The train/ and test/ sub-directories are consumed directly by ImageAI’s ModelTraining and CustomImagePrediction APIs. 
detect_kermit/
├── data/
│   └── images/
│       ├── train/
│       │   ├── kermit/kermit-train-images/
│       │   └── no-kermit/no-kermit-train-images/
│       ├── test/
│       │   ├── kermit/kermit-test-images/
│       │   └── no-kermit/no-kermit-test-images/
│       ├── models/
│       │   └── kermit_finder.h5
│       └── json/
│           └── model_class.json
├── helpers/
│   ├── convert_vid2image.py
│   ├── download_from_google.py
│   ├── rotate_images.py
│   └── utils.py
├── episode3_results/
├── imageai_build_model.py
├── kermit_model_evaluation.py
└── requirements.txt
The episode3_results/ directory is created manually (or expected to exist) before running video inference — the script writes annotated JPEG frames there at runtime.

​
model_class.json

ImageAI requires a class-map JSON file that translates integer class indices (as produced by the model’s softmax output layer) into human-readable label strings.
data/images/json/model_class.json
{
    "0" : "kermit",
    "1" : "no-kermit"
}
  • Index 0 → kermit — frames or images in which Kermit the Frog is detected.
  • Index 1 → no-kermit — frames or images in which Kermit is not present.
The path to this file is passed to the model at runtime via: 
model.setJsonPath(os.path.join(EXECUTION_PATH, 'data/images/json/model_class.json'))
If the JSON file is missing or the keys do not match what was used during training, ImageAI will raise an error when loadModel is called.

​
Model file (kermit_finder.h5)

PropertyValue
Filenamekermit_finder.h5
Locationdata/images/models/kermit_finder.h5
FormatHDF5 — Keras .h5 serialisation format
ArchitectureResNet, configured via setModelTypeAsResNet()
StorageTracked with Git LFS (Large File Storage) due to file size
The model is produced by running imageai_build_model.py and is subsequently loaded in kermit_model_evaluation.py with: 
model.setModelPath(os.path.join(EXECUTION_PATH, 'data/images/models/kermit_finder.h5'))
model.loadModel(num_objects=2)
To pull the .h5 file after cloning the repository, make sure Git LFS is installed: 
# macOS
brew install git-lfs

# Ubuntu / Debian
sudo apt-get install git-lfs
Then run git lfs pull from the repository root to download the model checkpoint.

​
requirements.txt

All Python dependencies are pinned to exact versions for reproducibility. Install them with: 
pip install -r requirements.txt
PackageVersion / SourcePurpose
pandas0.23.4Data manipulation utilities
imageai2.0.2 (GitHub wheel)Custom image prediction and model training API
pillow5.4.1Image I/O for rotation helpers
keras2.2.4Deep learning backend (used by ImageAI)
opencv-python4.0.0.21Video capture and frame annotation
scipy1.2.0Scientific computing (ImageAI dependency)
matplotlib3.0.2Plotting and visualisation
h5py2.9.0Reading and writing .h5 model files
tensorflow1.12.0ML framework (CPU build)
tensorflow-gpu1.12.0ML framework (GPU build)
google_images_downloadlatestDownloading additional training images from Google
imageai is installed directly from the GitHub release wheel rather than PyPI: 
https://github.com/OlafenwaMoses/ImageAI/releases/download/2.0.2/imageai-2.0.2-py3-none-any.whl
requirements.txt lists both tensorflow==1.12.0 and tensorflow-gpu==1.12.0. You should install only one of these depending on your hardware. Installing both in the same environment can cause import conflicts and unpredictable behaviour. Remove whichever line is not appropriate for your setup before running pip install -r requirements.txt.On macOS, the standard PyPI wheel for TensorFlow 1.12 does not work. Install it from the official storage URL instead:
python3 -m pip install --upgrade https://storage.googleapis.com/tensorflow/mac/cpu/tensorflow-1.12.0-py3-none-any.whl

​
EXECUTION_PATH

At the top of kermit_model_evaluation.py, the current working directory is captured once at import time: 
EXECUTION_PATH = os.getcwd()
This variable is used to build all file paths for the model checkpoint, the class JSON, and predicted image frames: 
model.setModelPath(os.path.join(EXECUTION_PATH, 'data/images/models/kermit_finder.h5'))
model.setJsonPath(os.path.join(EXECUTION_PATH, 'data/images/json/model_class.json'))
kermit_model_evaluation.py must be run from the repository root directory. If you invoke the script from a different directory, EXECUTION_PATH will not point to the data/ tree and the model will fail to load. Always use:
cd /path/to/detect_kermit
python kermit_model_evaluation.py -t image -f kermit.jpeg

Build docs developers (and LLMs) love

Get started for freeTalk to us