View Training Results

Training graphs and insights are made available for all Roboflow Train jobs.

During Training

After you start training a model, a message will appear on the dataset version page associated with the model you are training. This message will first say that a training machine is starting.

When training begins, a training graph will appear that shows changes in mAP during training over each epoch.

After Training

When your model has finished training, you can view metrics from the training process.

Object detection projects report on the precision, recall, and mAP associated with each training job in the Versions tab of your project.

Classification projects report on accuracy compared to a validation dataset, with a confusion matrix showing performance of the project.

Segmentation models show mAP score.

Click on "Details" next to the precision, recall, and mAP statistics on a project Version to see more granular information.

The "Details" tab shows a range of graphs showing performance of your training job. This information is useful if you want to dive deeper into your model performance.

Interactively Test Model Performance

See how your model performs on images from your dataset and local and hosted image and video files with the Deploy Tab.

You can also see how your model performed on the specific images in your test and validation sets by using the "Visualize" feature to get a subjective feel for what these training metrics mean for your particular dataset.

Last updated