MRI　(Magnetic　Resonance　Imaging)　is　a　powerful　tool　for　doctors　and　researchers　to　begin　diagnosing　brain　tumors　in　general.　After　the　MRI　detects　a　tumor,　a　common　way　to　determine　the　type　of　the　tumor　is　to　look　at　the　results　from　a　sample　of　tissue　after　a　biopsy　or　surgery.　However,　diagnosing　the　brain　tumor　type　based　on　the　sample　tissue　is　time-consuming　for　brain　tumor　experts.　There　could　also　be　some　misdiagnosis　if　the　brain　tumor　doctor　is　not　experienced　enough.　Therefore,　our　research　goal　is　to　use　deep　learning　technology　for　brain　tumor　detection　and　diagnosis　based　on　the　given　brain　MRI　image.　This　research　tried　different　deep　learning　models　to　test　their　efficiency　and　accuracy,　including　VGG16/19,　AlexNet,　GoogleNet,　Resnet.　We　also　implemented　a　YOLO　model　to　snip　the　tumor＇s　exact　location　in　the　MRI　image　and　make　graphical　enhancements　to　improve　classification　accuracy.　It　turns　out　that　the　classification　accuracy　after　using　YOLO　detection　and　graphic　enhancement　decreases.　Some　of　our　research　results　demonstrate　that　deep　learning　models　can　contribute　to　brain　tumor　diagnosis　to　a　great　extent,　improving　the　accuracy　and　diagnosis　speed,　so　it　is　possible　to　use　a　deep　learning　model　for　brain　tumor　diagnosis　and　further　complex　diseases　in　the　future.　©　2021　IEEE.