As part of my final year research project, I undertook the task of enhancing Google's EfficientDet, a state-of-the-art object detection model. By making specific modifications to the model architecture, I was able to improve its mean Average Precision (mAP) from 34% to 35%. This improvement was achieved through rigorous experimentation and training on Google’s Tensor Processing Units (TPUs) under the TensorFlow Research Program. The success of this project was recognized by the author of the official EfficientDet repository, who subsequently merged my code into the master branch, marking a significant contribution to the ongoing development of EfficientDet.