Open zht-ttt opened 3 years ago
以下是基于frcnn的 tensor([[-1.0611e-07, 7.5014e-07, 9.5363e-06, 1.5857e-05, 1.4884e-05, 5.9838e-06, 3.0337e-08], [ 1.4148e-06, 8.0262e-06, 4.2682e-05, 4.9330e-05, 4.3010e-05, 2.9354e-05, 1.7929e-06], [ 5.2836e-06, 3.0520e-05, 4.9705e-05, 4.5834e-05, 4.7702e-05, 4.3880e-05, 1.0087e-05], [ 5.4636e-06, 4.6103e-05, 7.7855e-05, 6.4577e-05, 5.9664e-05, 6.0945e-05, 1.3587e-05], [ 2.9276e-06, 3.3370e-05, 5.7708e-05, 4.8958e-05, 4.4900e-05, 4.6765e-05, 8.8209e-06], [ 5.3333e-06, 4.8228e-05, 5.6663e-05, 3.6995e-05, 2.9427e-05, 3.4944e-05, 1.0533e-05], [ 4.8740e-06, 1.6906e-05, 2.1360e-05, 1.5230e-05, 1.0777e-05, 1.2249e-05, 7.2816e-06]], grad_fn=)
以下是基于resnet50的 tensor([[-0.0170, -0.0125, 0.0583, 0.0920, 0.0787, -0.0123, -0.0159], [ 0.0004, -0.0086, 0.0673, 0.1415, 0.1246, 0.0116, -0.0072], [ 0.0158, 0.0534, 0.0699, 0.0970, 0.1079, 0.0683, 0.0301], [ 0.0197, 0.0577, 0.0239, 0.0880, 0.0655, 0.0634, 0.0259], [ 0.0156, 0.0606, 0.0234, 0.0019, 0.0289, 0.0436, 0.0418], [ 0.0097, 0.0523, 0.1047, 0.0886, 0.0741, 0.0924, 0.0857], [ 0.0128, 0.0616, 0.0922, 0.0549, 0.0591, 0.1060, 0.0816]], device='cuda:0', grad_fn=)
以下是基于frcnn的 tensor([[-1.0611e-07, 7.5014e-07, 9.5363e-06, 1.5857e-05, 1.4884e-05, 5.9838e-06, 3.0337e-08], [ 1.4148e-06, 8.0262e-06, 4.2682e-05, 4.9330e-05, 4.3010e-05, 2.9354e-05, 1.7929e-06], [ 5.2836e-06, 3.0520e-05, 4.9705e-05, 4.5834e-05, 4.7702e-05, 4.3880e-05, 1.0087e-05], [ 5.4636e-06, 4.6103e-05, 7.7855e-05, 6.4577e-05, 5.9664e-05, 6.0945e-05, 1.3587e-05], [ 2.9276e-06, 3.3370e-05, 5.7708e-05, 4.8958e-05, 4.4900e-05, 4.6765e-05, 8.8209e-06], [ 5.3333e-06, 4.8228e-05, 5.6663e-05, 3.6995e-05, 2.9427e-05, 3.4944e-05, 1.0533e-05], [ 4.8740e-06, 1.6906e-05, 2.1360e-05, 1.5230e-05, 1.0777e-05, 1.2249e-05, 7.2816e-06]], grad_fn=)
以下是基于resnet50的 tensor([[-0.0170, -0.0125, 0.0583, 0.0920, 0.0787, -0.0123, -0.0159], [ 0.0004, -0.0086, 0.0673, 0.1415, 0.1246, 0.0116, -0.0072], [ 0.0158, 0.0534, 0.0699, 0.0970, 0.1079, 0.0683, 0.0301], [ 0.0197, 0.0577, 0.0239, 0.0880, 0.0655, 0.0634, 0.0259], [ 0.0156, 0.0606, 0.0234, 0.0019, 0.0289, 0.0436, 0.0418], [ 0.0097, 0.0523, 0.1047, 0.0886, 0.0741, 0.0924, 0.0857], [ 0.0128, 0.0616, 0.0922, 0.0549, 0.0591, 0.1060, 0.0816]], device='cuda:0', grad_fn=)