The oil spill is a big ecological threat to oceans and coastal areas. An accurate detection of oil spills using images can help in timely action to reduce threat. This work uses a dataset gathered via drone technology to explore the usage of a convolutional neural network (CNN-based model for oil spill identification. The dataset, acquired between September 2021 and September 2023, covers a broad spectrum of environmental variables and operational settings in port environments. Images of oil spills were methodically labeled to distinguish them from non-oil spills things such as coastal buildings and water. Trained to recognize spatial hierarchies of characteristics, CNN could identify photos as either “Oil Spill” or “No Oil Spill.” With a precision of 0.90 for the categorization of oil spills and an area under curve (AUC) of 0.84, the model attained an overall accuracy of 0.83, therefore demonstrating great performance. The model's validation accuracy varied despite great training accuracy above 90%. The results imply that although the CNN can efficiently detect oil spills, consistent validation performance requires more fine-tuning and stability enhancement even if this is already possible.