OFEODM: An Onboard Fast and Efficient Object Detection and Distance Prediction Model

Object detection and distance prediction have advanced significantly in recent years. The YOLO series has released its 11th version, along with numerous variants that have been applied across various fields. Meanwhile, the Detection Transformer (DETR) has repeatedly set new state-of-the-art (SOTA) records in the field of object detection. Depth Anything also released its second version last year, further pushing the boundaries of distance detection. Although these models achieve impressive performance, they often require substantial computational resources. However, for algorithms intended for real-world applications and deployment on onboard devices, computational resources are limited and valuable. Inference time per frame is a critical factor in ensuring an algorithm's reliability and feasibility. Designing a model that operates in real time without sacrificing accuracy remains an extremely challenging problem, and extensive research is ongoing in this area. To address this challenge, we present a model called the OFEODM(Onboard Fast and Efficient Object Detection and Distance Prediction Model), which runs in real time on a comma.ai device equipped with a Qualcomm Snapdragon 845 processor. We deployed the comma.ai device on a car for autonomous driving purposes. Furthermore, experimental results of comparing our model to the state-of-the-art one-stage object detection models of the YOLO series indicate that our model demonstrates comparable performance but with faster speed on our collected real-world dataset. Additionally, we have incorporated an extra module into our Fast Detection model that enables it to predict the distance between our vehicle and detected objects, providing valuable information for downstream tasks such as path planning.