🔬 Data Science

Let's now leverage the created DatasetVersioncreated through the Data Management features to train the best model possible using the Data Science features of Picsellia.

Basically, the Picsellia Data Science part is basically a powerful and complex experiment tracking tool designed for Computer Vision

1. Leverage Project as an AI lab to perform as many Experiment as needed to create the best-performing ModelVersion

2. Organize and structure the development of your ModelVersion through Experiment to keep a clear and understandable track of all the work done by you and your collaborators

3. Train your ModelVersion either on the Picsellia Training engine or on your own computing resources

4. Leverage the Experiment Tracking dashboard to visualize Metrics helping you to assess the quality of the trained ModelVersion

5. Integrate your own model files and training script with Picsellia to, for instance, personalize the Metrics computed and displayed in the Experiment Tracking dashboard

6. Use the Evaluation Interface to appreciate the behavior of the ModelVersion on a dedicated bunch of images

7. Compare your Experiment to pick the best one

8. Export your best Experiment as a ModelVersion that will be stored for further usage (such a deployment) in your private & secured Model Registry.