Machine learning engineer to create ML dataset to test images and object detection

I need an app to detect pills from a large dataset of images. The goal would be to submit those images to train a ML. We can then use object detection to detect the images and have the trained data tell what that object is from the data it learned. Looking for the results to be - Make the full training library - ios app detects the object and displays/outputs the data it detected. 100k+ images to be learned, takes about 1-3 seconds for the trainer to read each image. I'm aware. There is a small dataset to use to test and create the app to test and demo. Then update the ML to read the remaining images I think would be the best method.