Previous page
Next page
Starting training
To start training with default parameters, do the following:
Click the Settings button in the Training section. The editable parameters will be displayed:
- Batch size—the number of images given to the neural network at one time.
- Learning rate—the speed at which the neural network adjusts its coefficients during training.
If necessary, you can modify additional parameters. To do this:
Add ?advanced=true to the end of the browser's address bar.
Press Enter. This will display an Advanced button and the following parameters:
- Epochs—the number of training cycles the neural network should complete.
- Accuracy—the desired accuracy level of the neural network (value range [70, 100]).
- Timeout—the maximum training duration in seconds (value range [5, 3600]).
Attention!
It is not recommended to change the default values of Batch size, Learning rate, and Epochs parameters unless necessary.
Click the START TRAINING button to initiate the process.
Attention!
Only one task can be trained at a time. Simultaneous training of multiple tasks is not allowed.
As a result, the training will begin, and the progress will be displayed next to the counter for completed epochs.
To stop the training process, click the STOP TRAINING button.
Retrieving results
After training is complete, in the Preview section, click the EXPORT MODEL button to download the neural network file.T he exported neural model is saved in the .annext format (supported in DetectorPack 3.13.1 and later versions). In the same area, in order to test object classification, you can upload the following files:
- a test image (by clicking the UPLOAD button);
- a frame from a webcam (by clicking the WEBCAM button).
The classification results will be displayed as a class match percentage on a scale of 0% to 100%.
To return to the task list, click the BACK button (see Web interface of TIY service).