The growth of tourism in the digital era encourages the use of social media as a source of visual data for destination analysis. This study aims to classify and map tourist attractions in West Kalimantan using a transfer learning-based Convolutional Neural Network (CNN) method with the MobileNetV2 architecture. A total of 454 images were collected through web scraping from the Instagram account @enjoykalbar, then through a process of elimination, augmentation, normalization, and manual labeling based on the West Kalimantan Disporapar tourism categories, namely Hills, Beaches, Cascades, Culture, Lakes, Rivers, Caves, and Forests. The dataset was divided into training data (70%), validation (20%), and test (10%). The model was built by freezing the initial layers of MobileNetV2 and adding a classification head, then drilled for 20 epochs using the Adam Optimizer and EarlyStopping and ReduceLROnPlateau callbacks. The training results showed a training accuracy of 95.8%, validation accuracy of 88.1%, and test accuracy of 80%. Further evaluation using the classification report yielded an overall accuracy of 89%, with an average precision of 0.93, a recall of 0.86, and an F1-score of 0.88. The model was then integrated into a category- and coordinate-based interactive mapping system to display the distribution of tourist attractions across 12 districts and 2 cities. The results demonstrate that the CNN transfer learning approach is effective for tourism image classification and supports spatial visualization in tourism promotion and planning.