the task#
Given an Image, Classify each pixel into 3 classes:
Segment Dog pixels.
Segment Cat pixel.
Segment Background pixels
venv:#
EnvImageSegmentation.yml
dataset#
Each image is an RGB image.
Image dimensions is non constant.
Masks are in the range [1, 2, 3]
01 prepare#
run 0001Dataset.py
Ensure the data in the folder Data.
label is black because it is 1,2,3 out of 256 which are all grey
if we draw by 3 colors only we will get:
for argmax we need 0,1,2 - convert to that ;
02 train#
help functions#
DL
├── DataLoader.py
├── Training.py
└── UNetModule.py