Advancing　swiftly　in　contemporary　society,　the　rapid　growth　of　autonomous　driving　technology　suggests　its　potential　adoption　across　continents.　The　realization　of　fully　autonomous　driving　relies　on　proficiently　detecting,　classifying,　and　tracking　road　objects　such　as　pedestrians　and　vehicles.　This　research　employs　the　YOLOv5　neural　network,　enhancing　it　with　YOLO-ALPHA.　Modifications,　encompassing　freeze　and　attention　mechanisms,　serve　to　refine　accuracy　and　expedite　training.　Furthermore,　adjustments　to　the　activation　function　aim　to　stabilize　precision　and　recall.　The　integration　of　an　FCN　based　on　semantic　segmentation　theory　contributes　to　improved　accuracy　in　detecting　road　conditions　during　autonomous　driving.　Consequently,　this　enables　the　successful　and　highly　accurate　functionality　of　automatic　identification.　©　The　2024　International　Conference　on　Artificial　Life　and　Robotics　(ICAROB2024).