Shrimp　is　a　type　of　aquatic　product　that　is　easy　to　deteriorate　and　the　freshness　has　an　essential　influence　on　both　its　taste　and　nutritional　value.　Scientists　have　developed　various　approaches　to　measure　shrimp　freshness;　however,　the　existing　methods　are　usually　destructive,　complicated　and　costly.　To　develop　a　fast,　non-destructive　and　low-cost　alternative,　we　utilized　deep　learning　models　to　identify　the　freshness　of　shrimp　based　on　photos　taken　by　smartphones.　The　models　were　trained　on　photographs　of　306　shrimp　along　with　their　total　volatile　basic　nitrogen　values　as　freshness　indicators.　Our　models　achieved　an　area　under　receiver　operating　characteristic　above　0.90　for　freshness　classification　and　root　mean　square　error　of　prediction　no　more　than　4.67 mg/100 g　on　fresh　samples　during　the　independent　tests.　Furthermore,　the　model　performance　was　evaluated　on　datasets　of　shrimp　photographed　for　7　consecutive　days　and　shrimp　placed　on　different　backgrounds　and　light　settings.　Our　study　suggested　deep　learning　as　an　accurate,　easy　and　low-cost　method　to　detect　shrimp　freshness,　which　may　have　broader　applications　in　food　safety.　©　2022,　The　Author(s),　under　exclusive　licence　to　Springer　Science+Business　Media,　LLC,　part　of　Springer　Nature.