Questions about

Model Development

Train and Test

Q: After training in the cloud platform, the test picture effect is very good, but the offline test picture effect is not good, why?

The image resolution of the online training model should be similar to that of the offline test. If you do not achieve the desired results when testing the image offline, please check that the resolution is the same.

Q: What is the training process and how long is it?

When the user clicks the "Training" button, the system will package the data (pictures, and annotated data) into training tasks and submit them to the server cluster, and the server cluster will arrange the training tasks. If too many tasks are being processed at the same time, new training tasks are queued.

After the training begins, the parameters of the deep learning network are constantly updated. This process takes about ten to twenty minutes, but it also depends on the number of training pictures and the training rounds (the system will adjust according to demand).

After the training is completed, the deep learning network will have the ability to detect such samples, which can be used for subsequent online testing and model export.

Still have a question?

Each section should be concise, user-friendly, and direct users to additional resources or documentation when necessary.