| Revision as of 06:40, 12 March 2024 editDB1729 (talk | contribs)Extended confirmed users, Pending changes reviewers, Rollbackers137,696 edits Category:Technology and applied science image insertion templates← Previous edit | Latest revision as of 22:47, 20 August 2024 edit undoPetri Krohn (talk | contribs)Autopatrolled, Extended confirmed users37,094 editsm Linked LeNet and AlexNet. | ||
| (One intermediate revision by one other user not shown) | |||
| Line 1: | Line 1: | ||
| + 1 = + 1 = 55. Since the kernel output is the same length as width, its area is 55×55.)]]<noinclude> | ] and ] convolution, pooling, and dense layers<br>(AlexNet image size should be 227×227×3, instead of 224×224×3, so the math will come out right. The original paper said different numbers, but Andrej Karpathy, the former head of computer vision at Tesla, said it should be 227×227×3 (he said Alex didn't describe why he put 224×224×3). The next convolution should be 11×11 with stride 4: 55×55×96 (instead of 54×54×96). It would be calculated, for example, as: + 1 = + 1 = 55. Since the kernel output is the same length as width, its area is 55×55.)]]<noinclude> | ||
| ] | ] | ||
| ] | |||
| </noinclude> | </noinclude> | ||
Latest revision as of 22:47, 20 August 2024
(AlexNet image size should be 227×227×3, instead of 224×224×3, so the math will come out right. The original paper said different numbers, but Andrej Karpathy, the former head of computer vision at Tesla, said it should be 227×227×3 (he said Alex didn't describe why he put 224×224×3). The next convolution should be 11×11 with stride 4: 55×55×96 (instead of 54×54×96). It would be calculated, for example, as: + 1 = + 1 = 55. Since the kernel output is the same length as width, its area is 55×55.)