It　is　challenging　for　most　existing　grippers　to　accurately　measure　their　contact　force　when　grasping　unstructured　objects.　To　address　this　issue,　a　novel　force　sensing　model　is　established.　A　compliant　gripper　derived　by　the　topology　optimization　method　is　introduced,　and　its　actual　deformation　is　measured　without　contacting　by　OpenCV.　Meanwhile,　the　hyperelastic　constitutive　model　of　flexible　materials　is　further　studied　by　the　uniaxial　compression　test　to　improve　the　accuracy　of　its　theoretical　deformation.　Subsequently,　the　force　sensing　model　is　established　based　on　linear　finite　element　theory　and　the　deep　neural　network　(DNN)　algorithm.　The　nonlinear　errors　of　actual　deformation　(input　layer)　and　theoretical　deformation　(output　layer)　are　compensated　by　the　DNN　algorithm.　This　compensated　deformation　is　then　input　into　the　linear　force　sensing　model　to　determine　the　contact　force.　Finally,　experimental　results　show　that　the　gripper　has　a　high　force　sensing　accuracy　(average　error　less　than　3%)　in　the　middle　part.　While　the　force　sensing　accuracy　at　the　end　of　the　compliant　gripper　has　declined,　the　contact　force　measurement　of　the　model　in　the　middle　of　the　new　compliant　gripper　has　been　effectively　verified.　©　2024　Author(s).