Machine learning life cycle
Follow the machine learning life cycle to generate a shoe classification model. The ML life cycle is different from the traditional software development life cycle. It consists of six specific steps:
- retrieve data
- Clean and prepare data
- Generate model
- Evaluation model
- Deployment model
- Obtain predictions (or inferences) from the model
The final result of the life cycle is a machine learning model that can be queried and returned with answers (or predictions).
Models are just mathematical representations of trends and patterns in the data. Good data is the foundation of all machine learning projects.
In step 1, get the data. In step 2, the data is cleaned, transformed, and placed in a format that can be learned by the machine. The cleanup and conversion process is usually the most time-consuming part of the machine learning life cycle. DJL allows developers to simplify this process by providing the ability to preprocess images using a translator. The translator can perform tasks such as resizing the image according to expected parameters or converting the image from color to grayscale.
Developers transitioning to machine learning often underestimate the time required to clean and transform data, so translators are a good way to start the process quickly. In the third step of the training process, the machine learning algorithm traverses the data multiple times (or over time) and then studies them to try to learn different types of footwear. The trends and styles found related to footwear are stored in the model. When evaluating the model to determine its ability to recognize footwear, step 4 is part of the training. If errors are found, correct them. In step 5, deploy the model to the production environment. After the model is put into production, step 6 allows the model to be used by other systems.
data
The footwear classification model is a multi-class classification computer vision (CV) model that uses supervised learning for training. The model classifies footwear into one of four category labels: boots, sandals, shoes, or slippers. Supervised learning must include numbers that have been marked with the goal (or answer) you want to predict