Hello!! I'm here to give some inputs to your problem. I've seen your sample image.
Being a research assistant in image processing and machine learning, my first intuition is that your problem is more of an image processing task than machine learning.
Here is how you can solve it:
1) We can convert and threshold the image into binary
2) We can erode and dilation the edges or boundaries that are darker than the wood piles frontal portion
3) We can count the resulting blobs or patches
4) We can find those patches that are circular or elliptical in shape
This should work pretty much everywhere
With machine learning, you will have to go through a different approach, though cannot guarantee a good solution. There shall be lot of false positives given that there are not much characteristics features of the frontal portion of the log piles. So, I'd recommend using the image processing way.
Software
MATLAB for testing
OpenCV for development
Python or C++
Then you can add the final numerical value of approximately found wooden logs to your app. You may add a button to click a picture, the output will show the logs marked in some color and the total count in that picture, in real time.
Thank you!!