Machine Learning Start-Up

Question

Ross MacAndrews 2017-4-19

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/336031-machine-learning-start-up

评论： Ross Mac Andrews 2017-7-4

I am needing to develop a deep learning program to identify and categorize various cuts of beef. I am currently collecting my own data set of photos, I should have about 1000 by the end, made up of maybe 5 or 6 different cuts that I need to identify.

can anyone offer advice on where to start and any recommendations, I am a complete beginner.

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Tish Sheridan 2017-5-24

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/336031-machine-learning-start-up#answer_268205

Hi - your use case sounds perfect for transfer learning - use your images to fine-tune a pretrained network to classify the cuts. Does this example help: http://uk.mathworks.com/help/nnet/examples/transfer-learning-and-fine-tuning-of-convolutional-neural-networks.html

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

Ross Mac Andrews 2017-7-4

Thanks very much for the reply,

Yes i have been looking at this example but am still confused. Am i correct in thinking the steps that I must take are:

1) Adjust the code to handle the dimensions of my images (I must first resize each of my images to be a consistent size)

2) Convert my images to black and white

3) apply to the example code.

Is using a network pre-trained on numbers or characters even going to benefit in identifying my images? (my images are of red meat (beef) cuts, vastly different from the numbers/characters) and as the example says "You can use transfer learning since the digits data is similar in content to the letters data." ???

All this considered, is there a better option i should be trying?

请先登录，再进行评论。