Train your model
Training a neural network model is a quite complicated process where one has to choose an architecture of the network, tune the parameters, choose the optimization method and so on. Luckily here we have done it for you!
Interactive model training is where you choose the type of a model and train it in one click. You will get a state-of-the-art model trained in just a few minutes! Moreover, if you an are advanced user, you can adjust many of the training parameters yourself.
In the basic view there are options to write the name of your model and set the training time. In object detection training you would have an additional field to choose when to stop training if the mean average precision does not improve for a certain duration. You will also see some information about your data:
- Count of images for each specific label
- Total number of images that will be used to train the model
- Total number of distinct labels
In the top left corner you can choose to switch to Advanced view, which enables you to change more parameters for the training session. Each parameter has an explanation which you can access by clicking the question mark next to it.
You can adjust the percentage of images to be used for model evaluation after training by modifying the validation set size. If you want to use manually chosen images for evaluation, check use user-defined validation set. To manually add images to the validation set, in the main page select the images you want to add, right click on one of them for the context menu to open and select add to validation set.
When you are done setting all the parameters, press Start to begin training. Trained models menu will open in which you will see the training process. During training you can work on other tasks.
Learn more about the process
What training time should I choose for a model?
That depends on the number and size of your images. We recommend selecting at least 3 minutes for single-label classification and at least 5 minutes for multi-label classification. Object detection model training takes much longer time, we recommend to set 360 min time limit and 60 minutes "no improvement" condition. That is the training will stop if the model's performance on validation set does not improve for 60 minutes.
After training, you can analyze the statistics and check the actual predictions on the images to see if they are satisfactory. If you would like to achieve a higher accuracy, try increasing the training time or the number of training images. Please note that, longer training does not always provide a more accurate model. Therefore, we continuously test the model's performance during training and additionally keep the best model (according to the validation error).
Which images the algorithm is trained on?
By default, the model is trained on all images that are properly labeled for the selected model type.
However, if you wish to train only on a certain group of images, you may filter them yourself by using the panel on the lower left or you can simply select the images by using the mouse before training.