Image Classification

1. Load images

To begin, we will load the images from an example dataset included in Piximi by pressing Open in the top left. Select Open example project > Open human U2OS-cells example project to get started. Alternatively, if you would like to load your own images, press Open image in the top right.

The images correspond to U2OS cells expressing arrestin-GFP, and a beta2-adrenergic receptor. Upon receptor stimulation arrestin-GFP is recruited to the plasma membrane and eventually endocytosed resulting in vesicle like structures.

Fig. 12 Open the U2OS example dataset

2. Categorize images

In the Categories sub-menu on the right hand side you can see that there are already 3 classes defined for the U2OS example project. Click on the label to the left of these categories to turn on/off (/) the display of these images. The classes are:

• Unknown

  • This represents the uncategorized images. Piximi will predict which class these images belong to later

• Negative Control (cytoplasmic GFP)

• Positive Control (vesicular GFP)

Fig. 13 Explore the category menu. Turn on/off (/) a particular category to show/hide only those images.

Single-click to select a 2-3 images from the unknown category that best fit the Negative Control category. Once selected, click Categorize in the top right and select Negative Control. Do the same for 2-3 Positive Control images.

3. Train model

Click on the Classifier button to open the sub-menu and then click on the Fit settings icon. Within the menu that opens, you can select various parameters to adjust model training. Open the Dataset Settings menu to find the Train percentage option. This value controls what percentage of the images you have annotated will be used to train the model in Piximi. The remainder will be used to test how well Piximi can classify images not previously seen. We will use the default for now.

Fig. 14 Open the classifier settings.

In the top right, press the to begin training. Piximi will now look at the training subset of the images you have annotated and try to learn what links the input image to a particular class. Then, Piximi will apply what it has learned by examining the testing subset of images and compare the models answers to the image class.

Fig. 15 Explore classifier settings and then press to begin training.

At the bottom of the fit settings page you will see two graphs update as Piximi trains the model which show the accuracy and loss of the model over incrementing epochs.

Fig. 16 Training history of a successful classifier model for the U2OS example dataset.

What is an epoch?

An epoch is a measure of how many times the entire training subset is studied by the deep learning model. As the number of epochs increases, the model optimizes itself to improve classification performance.

However, increasing the number of epochs does not necessarily lead to better results and can instead result in overfitting. Click here to read more about overfitting (needs link).

Accuracy is a measure of how well the model has performed and is calculated as the ratio between the number of correct predictions and the total number of predictions. In this case, Accuracy refers to the accuracy of the model at correctly determining the class of images in the training subset of images.

(1)\[Accuracy = \frac{\text{Number of correct predictions}}{\text{Total number of predictions}}\]

Validation Accuracy is the accuracy when the model examines the testing subset of the data.

Validation accuracy vs accuracy

If you notice that your Validation accuracy value decreases as epochs increase but your Accuracy measurement continues to increase, this means that your model is fitting to the training subset of data better but your model is losing the ability to accurately predict on new data.

This is a result of overfitting as your model begins to pick up features within your image, such as noise, that are not relevant to classification. In essence, overfitting is when the model memorizes the answer to a specific question, rather than determining the answer from scratch itself.

Loss is another metric that is calculated on the training and test subsets of data and are depicted as loss and validation loss, respectively. Loss represents a summation of the errors the model has made during classification.

You can now exit the Fit settings by clicking the in the top left.

Fig. 17 Exit the fit settings menu.

4. Predict classes for unlabelled data

Once your model has been trained you can press Predict to run the trained model on the unannotated data. Once an image has been classified you will see the color on the image thumbnail update to that particular class.

Fig. 18 Predict the class of unknown images using your trained model.

Click Evaluate to see in-depth metrics on how well the model performed.

See also

Link to our on guide, or a guide somewhere else. eg:

For further information on the evaluation metrics, see our evaluating a model guide.

Copyright

The BBBC016 images used here are licensed under a Creative Commons Attribution 3.0 Unported License by Ilya Ravkin.