This thesis explores how a drone-based system can assist indoor crime scene investigations by reducing the risk of contaminating critical evidence. The projectfocuses on solving four main challenges: Scanning and mapping the crime scene,detecting evidence, marking restricted zones on a 2D map, and generating a safepath for investigators to avoid stepping on evidence.A DJI Tello Drone was flown over a simulated crime scene while capturingdownward-facing images. These images were stitched together using OpenCV’spanorama functionality to create a 2D crime scene map. A custom-trained YOLOmodel was then applied to detect weapons and bloodstains, and red overlays weregenerated to mark no-go zones on the map. Using cost maps, a shortest pathalgorithm was used to calculate a safe path around these zones.The system was tested over 20 different mock crime scenes. Through the tests,the system managed to achieve an average accuracy of 82 %.A user study comparing the system-generated and human-drawn paths showedthat the AI consistently produces shorter paths while avoiding all evidence. Thiswas largy because the system ignored non-relevant objects such as bags and clothes,which human participants often choose to avoid, leading to longer paths.The result suggests that a low-cost, drone-based system could help improvecrime scene investigations by providing a rapid initial overview of the scene layout. While some limitations exist, such as drone drift and model misclassification,the system meets the defined requirements. It provides a promising foundationfor future tools to enhance safety, accuracy, and efficiency in forensic work.