search

haixiansheng / unet-keras-for-Multi-classification

48 stars 6 forks source link

How to organize the training dataset? #1

Open sunyongke opened 5 years ago

sunyongke commented 5 years ago

Hello,

Thank you share your project, it's very useful to me to train my multi-classes model. I have two classes book and bottle. I have labeled them with software Labeme. And got the mask images for each image.

I confused about where I should put these images in if I use your code to train a model. Which directory I should put my original images in, and which directory I should put the mask images in?

Thank you

haixiansheng commented 5 years ago

Hello  The original images put in train/image，local in code line 28. Mask images put in train/label, local in line 29.

发自我的iPhone

------------------ Original ------------------ From: sunyongke <notifications@github.com> Date: Tue,Nov 12,2019 4:05 PM To: haixiansheng/unet-keras-for-Multi-classification <unet-keras-for-Multi-classification@noreply.github.com> Cc: Subscribed <subscribed@noreply.github.com> Subject: Re: [haixiansheng/unet-keras-for-Multi-classification] How to organize the training dataset? (#1)

Hello,

Thank you share your project, it's very useful to me to train my multi-classes model. I have two classes book and bottle. I have labeled them with software Labeme. And got the mask images for each image.

I confused about where I should put these images in if I use your code to train a model. Which directory I should put my original images in, and which directory I should put the mask images in?

Thank you

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub, or unsubscribe.

sunyongke commented 4 years ago

Thank you for your reply. is the directory structure looks like as follow:

image--- |-class1 | |----a1.jpg |- class2 |----b2.jpg

haixiansheng commented 4 years ago

Two classes must be in one picture(mask), but labelme can mark two class with two different colors. Like:  floder1 : mask1(Class1, Class2), mask2(Class1, Class2),…… floder2: original picture1, original picture2, ……

------------------ Original ------------------ From: sunyongke <notifications@github.com> Date: Wed,Nov 13,2019 1:18 PM To: haixiansheng/unet-keras-for-Multi-classification <unet-keras-for-Multi-classification@noreply.github.com> Cc: haixiansheng <595956396@qq.com>, Comment <comment@noreply.github.com> Subject: Re: [haixiansheng/unet-keras-for-Multi-classification] How to organize the training dataset? (#1)

Thank you for your reply. is the directory structure looks like as follow:

image--- |-class1 | |----a1.jpg |- class2 |----b2.jpg

— You are receiving this because you commented. Reply to this email directly, view it on GitHub, or unsubscribe.