Home

Statistics Tool

The Statistics Tool is a powerful tool for evaluating the performance and accuracy of machine learning algorithms. By comparing the intermediate or final results of your models to ground truth data, you can gain valuable insights into their effectiveness and make informed decisions based on the results.

The tool provides visualizations, metrics, and statistical analyses to facilitate the comparison process. You can analyze accuracy, precision, recall, F1-score, or any other relevant performance metric for each model version, as well as user-defined segmentations. By using the Statistics Tool for model evaluation, you can make informed decisions about model enhancements, fine-tuning, and deployment strategies based on a comprehensive analysis of the data.

The Model Version Comparison feature within the Statistics Tool is particularly useful for tracking the progress and performance of a model over time or when experimenting with various iterations of the model. By comparing the results of different algorithm versions, you can gain valuable insights into the effectiveness of each version and make informed decisions about which version to use for a given task.

Finally, the Statistics Tool provides users with the ability to create preconfigured report suites that define how to evaluate their machine learning algorithms. These report suites can be customized to include different preconfigured views, which allow users to evaluate their models based on different metrics and segmentations over the test data. By using preconfigured views, users can easily identify which algorithms are performing well and which ones need improvement, allowing them to make informed decisions about how to optimize their models. Additionally, the creation of a report suite can be particularly useful for users who need to share their results with others or who want to keep track of their model's performance over time.

The tool has several user-defined functions that can be implemented to customize it for specific project characteristics. The Reading function is used to parse the prediction log files and the ground truth annotations and return them as a dataframe. The Association function is used to match different results in the same inference while the Transform function is optional and used to reshape the dataframe with the data.

In the report viewing phase, there are additional user-defined functions available. The Metric function can be used to calculate various metrics about the accuracy of the algorithm's predictions. The Partitioning function can be used to partition the data based on specific criteria. The Statistics function can be used to calculate statistics about the accuracy of all the samples, and the Confusion function can be used to calculate the accuracy of the algorithm's predictions for each row in the partitioned data.