I make a lot of stuff, like a lot. This time, I made a script with OpenCV and Tensorflow, two of the most advanced deep learning libraries.
I was able to make a script in Python which detects about 130 common objects, like chairs, tables, cups, and so on in an image. The script uses OpenCV with Tensorflow. For a long time I have always been against Tensorflow, but lately my views on it have changed significantly lately. The combination of Tensorflow and OpenCV works incredible.
Now while my results still aren't perfect, it is better than AWS. OpenCV and Tensorflow were able to identify a few more objects in the image, while most were accurately identified, some were not properly identified, like the umbrella being seen as a person. In my case, MATPLOTLIB was also able to help alter the color of things, like altering skin color to see where the computer sees the difference in objects. There is also another benefit to this; when images are more overexposed (meaning the coloring is brighter), computers are able to identify differences a lot easier and be more accurate in results.
The datasets being used were YOLOv3. It is identifying 130 of the most common objects.
This was just a quick, short post I wanted to make on my deep learning adventures. Hopefully this provides a little insight!