Image Classification Example

Login

1. Launch up the application and log in.

Project Creation

1. Create a new project by filling in all the required info and select datasets.
2. If a dataset is not available, select the option create dataset/upload labelled dataset.
3. If create dataset is selected, the dataset will require manual labelling later.
4. If upload labelled dataset is selected, either a labelled dataset can be uploaded or unlabeled dataset can be labelled automatically by the app.

Dataset Creation/Labelling

Dataset Creation

1. Select create dataset.
2. Select webcam for capturing images, select either webcam or IP cameras and start the camera.
3. Click start capturing and navigate to the file directory shown in the app to locate the images.

4. Next, select file upload and upload the zipped images and give the dataset a name.

Upload Labelled Dataset

1. To upload a labelled dataset, zip the images with its labels and upload into the app.
2. To use the app to perform multiple labels, upload in a file format like shown here File.
3. By uploading this format, the app will label all the images with the folder name.

Labelling

1. This section can be skipped if a labelled dataset was uploaded and selected.
2. If the uploaded dataset is unlabeled, proceed to this section and perform the labelling.
3. The app provides a Label-Studio app UI for labelling.
4. First, add in the label that is desired and press Enter.

5. Save to update the labels.
6. Next, proceed to labelling all the desired data by selecting Start Labelling at the top of the page.

Training

1. After dataset labelling, proceed to train a model.
2. Create a training session and select a pre-trained model or user-trained model.

3. Once a model is selected, configure the training session by adjusting parameters such as epochs, optimizer, batch size, augmentations and etc.
4. Start the training and wait for it to finish.
5. Adjust the parameters and continue to training if necessary. (Beware: Be sure to allow at least one training to complete to allow the app to save the model for continue training)

Deployment

1. Select the desired model to deploy.
2. Select the desired media for deployment. (e.g. image, video, mqtt)
3. If necessary, it is also possible to upload a well train model in this section using the options at the sidebar.