🧘Yoga Pose Image Classification with Pretrained Models

This project is a small introduction to Computer Vision! In this example, we will re-train an pre-trained model to recognize yoga poses from image data.

🛠️ Installation

Clone the GitHub repository to your local machine:

git clone https://github.com/vriveraq/Yoga-Pose-Image-Classification.git

Create an environment and install your requirements!

Clone the GitHub repository to your local machine:

pip install -r requirements.txt

👣 Pipeline

We will follow the general machine learning workflow.

Examine and understand the data
Build an input pipeline, in this case using Keras ImageDataGenerator
Compose the model
Load in the pre-trained base model (and pre-trained weights)
Stack the classification layers on top
Train the model
Evaluate model

📈 Data

Yoga is a very well-known practice to curb anxiety and relieve stress. There are many yoga poses but the very well-known ones are the downward dog pose, goddess pose, tree pose, plank pose and warrior pose.

Computer Vision can be used to classify between them!

🔎 Caveat: The dataset is divided into train and test subdirectories, with 5 subfolders in each directory corresponding to the 5 classes of yoga poses. The images are extracted from Bing using their API functionality so they may not be very accurate (watermarks, text may exist).

This dataset was obtained from Kaggle.

🤳Model: VGG19

There are many pre-trained models for image classification! You can see a full list here: https://keras.io/api/applications/. We will use the VGG-19 model (read more).

Notes:

For image classification use cases, see this page for detailed examples.
For transfer learning use cases, make sure to read the guide to transfer learning & fine-tuning.

Model features:

The default input size for this model is 224x224.
Each Keras Application expects a specific kind of input preprocessing. For VGG19, call tf.keras.applications.vgg19.preprocess_input on your inputs before passing them to the model. vgg19.preprocess_input will convert the input images from RGB to BGR, then will zero-center each color channel with respect to the ImageNet dataset, without scaling.

📚 Resources

You can follow this complete tutorial applied to a simpler scenario building a cat and dog classifier: Tensorflow: Transfer learning and fine-tuning.

🌐 Contact details

Find this project interesting. Let's connect! LinkedIn: vriveraq

vitaly-shalem/Yoga-Pose-Image-Classification