Demand for automated image classification is spiking every day. This study aims to develop an Inception-V3-based classifier on soil nitrogen to facilitate the immediate determination of nitrogen present in the soil. To determine the Nitrogen features, the study used a Rapid Soil Test Kit color chart from the Bureau of Soils and Water Management. To develop a Nitrogen image classifier, the study used Inception-v3 in which was first trained on ImageNet and was repurposed to learn Nitrogen features. Specifically, the method of learning is transfer learning. To determine the Nitrogen Image Classifier’s usability in the field test trial, samples were taken from the farm. The findings revealed that the representatives verified the image classifier’s prediction by manually comparing the soil test result to the color chart, making judgments, and comparing the image classifier prediction to their judgement. The soil nitrogen image classifier prediction trial was set at 5 trials. Colors can be used to train an image classifier. They are a reliable source of information to identify images. 20 images of different angles per label are enough to train the Inception-v3 model. The retrained model has a 94.8846% prediction accuracy on soil nitrogen.