Object classification models are essential to computer vision applications but they frequently include a large number of parameters and lacks adaptiveness towards unseen data. This project proposes a modern object classification architecture with reduced parameters by replacing traditional fully connected dense layers with a Liquid Neural Network (LNN). The proposed architecture will comprise several Convolutional Neural Network (CNN) layers connected sequentially to extract spatial features from images, followed by a liquid neural network to process the extracted features. The LNN’s special time-dependent adaptive properties allow the model to respond dynamically to new, unseen data beyond the training set. Finally, to classify the objects in the image, the output of the LNN layer is passed through a SoftMax activation function. This design seeks to achieve great parameter efficiency while improving robustness and adaptability by substituting LNNs for conventional dense layers. The study will also investigate ways to further enhance object classification performance by utilizing the flexibility of the LNN.