(Trying my hands on Deep Learning)
-
Build an "Image Classifier" using
Python
,Tensorflow
andKeras
. -
I have created this Image classifier in a
virtual environment
which is a good and ideal practice while working on these kind of projects as we import lots of libraries and dependies so we don't want those files to mess up with other projects. -
I have used command :
!pip install tensorflow==2.8 tensorflow-gpu==2.8 opencv-python matplotlib protobuf==3.19.1
instead of!pip install tensorflow tensorflow-gpu opencv-python matplotlib
as Tensorflow has dropped native windows support for GPU usage in their latest versions of tensorflow -
1 and 0 labels done by the dataset API relate itself to the 0 and 1 determined by the model's sigmoid function
(0 - Happy , 1 - Sad). -
Got a handful experience on using
cuDNN
andCUDA
, In particular the versions I have used are given below : -
CUDA
,cuDNN
andMSVC
version should be carefully selected according to the tensorflow version in you system otherwise It wont work properly !
-
We use
tensorflow-gpu
instead of simpletensorflow
to reduce the time taken by model in training as we generally have a large dataset so It takes a lot of time that's where GPU helps in making process a lot faster.
(You can use!pip install tensorflow opencv-python matplotlib
for tensorflow without GPU) -
Can we run without GPU ? Answer is YES !