On March 27, 2018, a four-liter bottle of dichloromethane was spilled in a lab at Texas Tech University. The unnoticed spill exposed researchers to high vapor concentrations, causing poisoning and hospitalization. This incident emphasizes how dangerous an undetected liquid spill can be in a controlled environment like a laboratory, which, at its worst, can lead to containment breaches and life-threatening situations. However, the problem is not confined to research labs. Unknown liquid and chemical spills are common in manufacturing facilities, hospitals, military bases, and even public areas. To reduce this risk, this project developed a prototype liquid-detecting computer vision model. To generate masks, bounding boxes, and labels, the model used the Mask R-CNN architecture, which extends Faster R-CNN by adding a parallel mask prediction branch. Using training functions from Meta’s Detectron2 library, the model was trained for 12,500 iterations at learning rates of 0.01 and 0.001, resulting in a segmentation mask precision of 45.17 and a bounding box precision of 47.21. The fine-tuned model detected liquid spills and distinguished them from other objects with moderate accuracy, underscoring the potential of computer vision as an early-warning system for hazardous spills and a means to ensure safety.