Tank Classifier

This system uses deep learning to classify tanks and artillery vehicles.

For instructions, click the button below to open up the tutorial

Select an image containing a tank using the input below

Tank Classifier Tutorial

We have created a user-friendly tutorial to help users effectively utilize the AI-based tank classification model. The interface is easy to navigate, users can upload an image of a military vehicle, such as a main battle tank or artillery vehicle, by clicking the "Choose File" button. After selecting the image, the user can run the classification by clicking the "Classify" button, and the model will display the predictions as illustrated in the image below.

Upon analyzing an image, the tank classifier presents the user with a list of the 5 most likely variants, ranked by probability, with the highest score being the most probable match for the image; details about the predicted military vehicle; and additional images for comparison to help the user make a more informed decision, as the AI model may not always make the correct prediction.

It is important to note that the AI model will perform more accurately if the image is cropped to only include the specific vehicle being classified.

For example, look at the example image below,

If the image provided by the user shows a tank that is at a distance and partially obscured by the background/foreground elements, the AI model is likely to make an incorrect classification, as demonstrated in the example below.

The image shows a T-55 tank, but the model incorrectly identifies it as a K2 Black Panther. However, when the image is cropped and the AI model is re-run, it correctly classifies the tank as a T-55.

Additionally, the AI model was trained using only images of pristine tanks, meaning the tanks in the images were not damaged. However, the model still has a good ability to classify destroyed tanks correctly in most cases. Please see the 2 images below.

Even though the images contain destroyed tanks, the AI model still correctly classifies in both instances.

Please note that this model only works for the following 20 classes of the military vehicles (18 main battle tanks, and 2 artillery vehicles).

['Al Khalid', 'Arjun', 'Armata', 'Challenger 2', 'K2 Black Panther', 'K9 Thunder', 'Leclerc', 'Leopard 2', 'M109', 'M1A2 Abrams', 'Merkava Mk.4', 'Oplot-M', 'T-55', 'T-72', 'T-80', 'T-90', 'Type 74', 'Type 90', 'Type 99', 'VT-4']

The training dataset is not large as well i.e., only around 110 images were used for each of the category to train the models. If the model is trained using more data, it will become more better at classifying.