A pedestrian and vehicle detection algorithm CFAI-YOLO based on YOLOv11s is proposed to process the issues of insufficient accuracy, high false detection and missed detection rates in existing pedestrian and vehicle detection algo-rithms. Firstly, the introduction of context and spatial feature calibration networks effectively solves the problems of pixel context mismatch and spatial feature misalignment. Secondly, an adaptive fine-grained channel attention mechanism is added to the backbone network to achieve more effective feature weight allocation. Finally, using the ADown downsam-pling module preserves the target feature information and reduces the number of model parameters. The experimental of the improved model on the KITTI dataset achieved 88.1% mAP, which is 2.4% higher than the original model.