Adding data to an existing recipe and retraining

Adding Images for Training

Step 1 :

To connect to the network cable, first, locate the Ethernet port on your camera and connect one end of the network cable to it. Then, connect the other end of the cable to your router or switch. To log into the camera, you can open a web browser on your computer and enter any of the given IP addresses of your camera into the address bar. This should take you to the login page of your camera.

Step 2 :

To access the library, look for the library button on your software's menu bar and click on it. Once you're in the library, you can select the recipe you want to view by filtering for it.

To filter the images by fail, look for the filter option and select 'fail'. This will display all the failed images in the recipe. You can further organize them by sorting them by day, date, and time by selecting the appropriate sorting option.

Step 3 :

To add all the images in the failed section that are not supposed to fail, you can first select them all by holding down the 'Ctrl' key on your keyboard and clicking on each image you want to select. Once you have selected all the images, you can look for the 'Add to active recipe's trainset' button and click on it. This will add the selected images to the active recipe's trainset, which will help improve the accuracy of your model. It's important to only add images that are not supposed to fail to your trainset, as adding failed images can negatively impact the performance of your model. Once you have added the selected images to the trainset, you can continue to train your model and test it on new data to see if the accuracy has improved.

Step 4 :

Clicking on this button or link will take you directly to the recipe editor, where you can make adjustments to your recipe and continue training your model. If you're having trouble finding the 'Go to Recipe Editor' button or link, refer to the software's user manual or contact the software's support team for assistance.

If a popup does not show up automatically after you add the images to the trainset, you can look for a button or link that says 'Go to Recipe Editor.' This button or link should be located on the same page where you added the images to the trainset.

Step 5 :

Once on the recipe editor page, select “View all ROIs”. To filter the images by class and select unlabeled, you can look for the filter option and select 'Class' and then 'Unlabeled'. This will display all the images in the recipe that have not been labeled yet. Once you have the list of unlabeled images, you can select all the images that you previously selected by holding down the 'Ctrl' key on your keyboard and clicking on each image you want to select. After selecting the images, look for the 'Label selected ROIs' button and click on it. This will bring up a dialog box where you can label the selected ROIs (regions of interest) in the images. You can choose the appropriate label for the selected ROIs and click 'OK' to apply the label to all the selected ROIs in all the selected images. This will help you organize your data and train your model more effectively.

Note

Take care when using the bulk labeling tool to avoid accidentally mislabeling data. Click the "clear selection" button at the bottom of the tool every time you re-label. If you fail to do so, the images might move over but remain selected, which can lead to incorrect re-labeling.

Step 6 :

When the dialog box appears after clicking on 'Label selected ROIs', you can choose the appropriate classification label for the selected ROIs. In this case, you have labeled the ROIs as 'False', but you can choose any appropriate label that fits your use case. To do this, select the label you want to choose from the 'Classification' drop-down menu. After selecting the appropriate label, click 'OK' to apply the label to all the selected ROIs in all the selected images. A confirmation dialog box will appear asking you to confirm the label selection, click 'OK' again to apply the label.

Step 7 :

Train the classification model - This will start the training process, where the camera will learn from the labeled data and adjust its algorithms to improve the accuracy of its predictions. Once the training process is complete, you can test the accuracy of the model by feeding it new data and observing its predictions. If the accuracy is not satisfactory, you can repeat the training process with additional data or adjust the model parameters to improve its performance.