Few-shot　object　detection　(FSOD)　shines　in　identifying　novel　category　targets　with　limited　support　samples,　making　it　well-suited　for　scenarios　with　restricted　data　availability.　However,　FSOD　suffers　challenges　in　acquiring　indicative　discrimination　features　due　to　the　limited　availability　of　support　samples　and　the　substantial　variations　in　object　shapes,　sizes,　and　distributions.　To　tackle　this　problem,　we　present　a　novel　few-shot　object　detector　with　fine-grained　support　information　guidance　based　on　Meta　R-CNN.　A　support-commonalities　enhancement　module　is　proposed　to　highlight　the　commonality　of　support　sample　features　channel-wise　and　preserve　spatial　information　by　external　attention　mechanism　to　extract　more　representative　category　features　of　support　samples.　Furthermore,　a　fine-grained　aggregation　module　is　designed　to　optimize　pixel-wise　spatial　alignment　of　information　between　support　and　query　features　for　generating　high-quality　proposals.　In　addition,　the　classification　compensation　module　aggregated　the　enhanced　support　information　in　classification　branches　to　enhance　the　model＇s　classification　accuracy　for　the　proposed　objects.　Finally,　the　proposed　detector＇s　effectiveness　is　validated　through　ablation　and　comparison　experiments　performed　on　the　MS　COCO　datasets.　　©　2024　ACM.