I’ve written code for mnist dataset and im getting 96% accuracy too. But when i input an image from my system and then try predicting images it always either predicts 6 or 8. I dont know what to do. I’ve tried inverting color too. Please help.
Hi, could you show one example of your own image here? and the tensor value of it.
Image=
Tensor representation=
tensor([[[0.7647, 0.7725, 0.7725, 0.7686, 0.7686, 0.7686, 0.7647, 0.7608,
0.7686, 0.7176, 0.7725, 0.7843, 0.7216, 0.7608, 0.8000, 0.7176,
0.7529, 0.7569, 0.7490, 0.7333, 0.7412, 0.7569, 0.7569, 0.7412,
0.7333, 0.7333, 0.7333, 0.7333],
[0.7647, 0.7647, 0.7686, 0.7725, 0.7686, 0.7647, 0.7608, 0.7569,
0.7490, 0.7961, 0.7490, 0.7451, 0.7843, 0.6706, 0.6118, 0.7490,
0.7490, 0.7529, 0.7451, 0.7333, 0.7333, 0.7529, 0.7529, 0.7412,
0.7373, 0.7373, 0.7333, 0.7333],
[0.7686, 0.7647, 0.7686, 0.7686, 0.7647, 0.7647, 0.7608, 0.7569,
0.7765, 0.7176, 0.7333, 0.7804, 0.6706, 0.5020, 0.5216, 0.6902,
0.7451, 0.7529, 0.7490, 0.7333, 0.7294, 0.7412, 0.7451, 0.7333,
0.7294, 0.7294, 0.7294, 0.7294],
[0.7647, 0.7686, 0.7647, 0.7647, 0.7647, 0.7608, 0.7569, 0.7529,
0.7451, 0.7412, 0.7529, 0.6510, 0.5216, 0.5255, 0.5882, 0.5686,
0.7373, 0.7529, 0.7569, 0.7373, 0.7333, 0.7412, 0.7412, 0.7333,
0.7294, 0.7294, 0.7294, 0.7294],
[0.7608, 0.7608, 0.7647, 0.7647, 0.7608, 0.7608, 0.7569, 0.7529,
0.7059, 0.8510, 0.6863, 0.4392, 0.5647, 0.7333, 0.6431, 0.4941,
0.7255, 0.7451, 0.7529, 0.7373, 0.7255, 0.7333, 0.7333, 0.7294,
0.7255, 0.7255, 0.7255, 0.7255],
[0.7647, 0.7647, 0.7647, 0.7608, 0.7608, 0.7569, 0.7529, 0.7569,
0.7961, 0.6392, 0.5098, 0.5451, 0.6902, 0.7804, 0.6667, 0.4627,
0.7176, 0.7412, 0.7529, 0.7412, 0.7294, 0.7333, 0.7412, 0.7373,
0.7255, 0.7255, 0.7255, 0.7255],
[0.7569, 0.7608, 0.7608, 0.7608, 0.7608, 0.7608, 0.7529, 0.7490,
0.7333, 0.4235, 0.5176, 0.7686, 0.7412, 0.7216, 0.6824, 0.4471,
0.7020, 0.7294, 0.7490, 0.7373, 0.7255, 0.7294, 0.7373, 0.7373,
0.7176, 0.7176, 0.7176, 0.7176],
[0.7608, 0.7647, 0.7608, 0.7608, 0.7608, 0.7608, 0.7529, 0.7529,
0.4471, 0.5608, 0.7255, 0.7686, 0.7333, 0.7725, 0.6902, 0.4392,
0.6980, 0.7294, 0.7451, 0.7373, 0.7216, 0.7294, 0.7412, 0.7412,
0.7176, 0.7176, 0.7176, 0.7216],
[0.7176, 0.7647, 0.7686, 0.7412, 0.7686, 0.7804, 0.6667, 0.5020,
0.5765, 0.7059, 0.7725, 0.7686, 0.7373, 0.7725, 0.7059, 0.4196,
0.6902, 0.7451, 0.7333, 0.7490, 0.7451, 0.7255, 0.7608, 0.7294,
0.7647, 0.7137, 0.6980, 0.7216],
[0.8118, 0.7569, 0.7412, 0.7725, 0.7529, 0.6431, 0.5294, 0.4863,
0.7059, 0.7490, 0.7490, 0.7373, 0.6941, 0.7098, 0.6902, 0.4824,
0.6157, 0.6784, 0.6471, 0.6353, 0.6039, 0.5569, 0.5686, 0.5333,
0.4902, 0.4784, 0.5098, 0.5882],
[0.7333, 0.7216, 0.7569, 0.7686, 0.6549, 0.5255, 0.5569, 0.6824,
0.7216, 0.6902, 0.6275, 0.5961, 0.5137, 0.4824, 0.4667, 0.3098,
0.3569, 0.4706, 0.4706, 0.4784, 0.4902, 0.4745, 0.5059, 0.4863,
0.5569, 0.5529, 0.5922, 0.6627],
[0.7451, 0.7804, 0.7451, 0.5765, 0.3882, 0.3176, 0.3725, 0.4510,
0.4118, 0.4275, 0.4471, 0.5098, 0.5176, 0.5608, 0.5961, 0.4549,
0.5451, 0.7020, 0.7137, 0.7059, 0.7255, 0.7098, 0.7255, 0.7216,
0.7490, 0.7255, 0.7176, 0.7333],
[0.7490, 0.7490, 0.6784, 0.5882, 0.5882, 0.6667, 0.6941, 0.6588,
0.6863, 0.7294, 0.7451, 0.7647, 0.7373, 0.7608, 0.7490, 0.5569,
0.5373, 0.7451, 0.7569, 0.7333, 0.7529, 0.7294, 0.7294, 0.7255,
0.7529, 0.7255, 0.7098, 0.7098],
[0.7725, 0.7333, 0.6980, 0.7059, 0.7412, 0.7686, 0.7608, 0.7373,
0.7176, 0.7490, 0.7451, 0.7490, 0.7137, 0.7451, 0.7412, 0.5529,
0.4275, 0.6902, 0.7216, 0.6941, 0.7451, 0.7333, 0.7333, 0.7451,
0.7137, 0.7137, 0.7176, 0.7255],
[0.7490, 0.7412, 0.7608, 0.7804, 0.7647, 0.7333, 0.7373, 0.7725,
0.7804, 0.7569, 0.7412, 0.7686, 0.7373, 0.7373, 0.7490, 0.6078,
0.4431, 0.7373, 0.7686, 0.7176, 0.7608, 0.7373, 0.7176, 0.7294,
0.7020, 0.7098, 0.7216, 0.7216],
[0.7686, 0.7608, 0.7451, 0.7412, 0.7412, 0.7451, 0.7373, 0.7294,
0.7725, 0.7216, 0.7020, 0.7804, 0.7529, 0.7255, 0.7412, 0.6471,
0.3765, 0.6941, 0.7333, 0.6863, 0.7412, 0.7294, 0.7176, 0.7373,
0.7176, 0.7255, 0.7294, 0.7176],
[0.7569, 0.7529, 0.7529, 0.7490, 0.7451, 0.7412, 0.7373, 0.7373,
0.7333, 0.7490, 0.7412, 0.7216, 0.7412, 0.7608, 0.7294, 0.6667,
0.3608, 0.7176, 0.7137, 0.7176, 0.7686, 0.7020, 0.7333, 0.7216,
0.7137, 0.7137, 0.7137, 0.7176],
[0.7529, 0.7490, 0.7490, 0.7451, 0.7412, 0.7373, 0.7333, 0.7333,
0.7333, 0.7451, 0.7373, 0.7216, 0.7373, 0.7529, 0.7216, 0.6627,
0.3725, 0.7098, 0.7137, 0.7137, 0.7569, 0.7020, 0.7333, 0.7216,
0.7137, 0.7137, 0.7137, 0.7176],
[0.7451, 0.7451, 0.7412, 0.7373, 0.7373, 0.7333, 0.7333, 0.7294,
0.7333, 0.7412, 0.7333, 0.7176, 0.7294, 0.7412, 0.7137, 0.6627,
0.3922, 0.6980, 0.7176, 0.7137, 0.7412, 0.7020, 0.7294, 0.7216,
0.7137, 0.7137, 0.7137, 0.7176],
[0.7373, 0.7373, 0.7373, 0.7333, 0.7333, 0.7294, 0.7294, 0.7294,
0.7333, 0.7373, 0.7294, 0.7176, 0.7255, 0.7294, 0.7020, 0.6667,
0.4118, 0.6706, 0.7176, 0.7098, 0.7255, 0.7059, 0.7216, 0.7216,
0.7098, 0.7098, 0.7098, 0.7137],
[0.7333, 0.7333, 0.7333, 0.7333, 0.7333, 0.7294, 0.7294, 0.7294,
0.7333, 0.7333, 0.7255, 0.7216, 0.7216, 0.7216, 0.7020, 0.6745,
0.4235, 0.6431, 0.7216, 0.7098, 0.7137, 0.7098, 0.7176, 0.7216,
0.7098, 0.7098, 0.7137, 0.7098],
[0.7333, 0.7333, 0.7333, 0.7333, 0.7333, 0.7333, 0.7333, 0.7333,
0.7294, 0.7294, 0.7255, 0.7216, 0.7216, 0.7176, 0.7020, 0.6902,
0.4275, 0.6118, 0.7216, 0.7216, 0.7098, 0.7216, 0.7176, 0.7176,
0.7059, 0.7059, 0.7098, 0.7059],
[0.7333, 0.7333, 0.7333, 0.7373, 0.7373, 0.7373, 0.7373, 0.7373,
0.7294, 0.7255, 0.7216, 0.7255, 0.7255, 0.7176, 0.7098, 0.7059,
0.4275, 0.5843, 0.7176, 0.7216, 0.7098, 0.7255, 0.7176, 0.7137,
0.7098, 0.7098, 0.7098, 0.7059],
[0.7373, 0.7373, 0.7373, 0.7373, 0.7373, 0.7373, 0.7373, 0.7373,
0.7255, 0.7216, 0.7216, 0.7294, 0.7294, 0.7176, 0.7137, 0.7176,
0.4196, 0.5686, 0.7216, 0.7255, 0.7137, 0.7333, 0.7176, 0.7137,
0.7098, 0.7098, 0.7098, 0.7059],
[0.7294, 0.7294, 0.7294, 0.7255, 0.7255, 0.7216, 0.7216, 0.7216,
0.7216, 0.7255, 0.7255, 0.7255, 0.7255, 0.7216, 0.7176, 0.7137,
0.4235, 0.5569, 0.6941, 0.7373, 0.7216, 0.7020, 0.7059, 0.7137,
0.7020, 0.7020, 0.6980, 0.6980],
[0.7294, 0.7294, 0.7255, 0.7255, 0.7255, 0.7216, 0.7216, 0.7216,
0.7216, 0.7255, 0.7255, 0.7255, 0.7255, 0.7216, 0.7176, 0.7137,
0.5333, 0.6157, 0.7059, 0.7294, 0.7176, 0.7137, 0.7098, 0.7020,
0.6980, 0.6980, 0.6980, 0.6980],
[0.7294, 0.7255, 0.7255, 0.7255, 0.7216, 0.7216, 0.7216, 0.7216,
0.7216, 0.7255, 0.7255, 0.7255, 0.7255, 0.7216, 0.7176, 0.7137,
0.6627, 0.6902, 0.7098, 0.7137, 0.7176, 0.7216, 0.7098, 0.6941,
0.6980, 0.6980, 0.6980, 0.6980],
[0.7255, 0.7255, 0.7255, 0.7216, 0.7216, 0.7216, 0.7176, 0.7176,
0.7216, 0.7255, 0.7255, 0.7255, 0.7255, 0.7216, 0.7176, 0.7137,
0.7216, 0.7176, 0.7137, 0.7098, 0.7137, 0.7137, 0.7098, 0.7020,
0.6980, 0.6980, 0.6980, 0.6980]]])
Predicted probabilities=
tensor([[ 2.4786, -3.3879, 1.1856, 5.9299, -2.3882, -2.1490, 0.2966, -8.1952,
12.4393, -8.1589]], grad_fn=)