Most of cloud services only offer you the possibility to detect a pre-defined set of objects. This might be enough for your business but in most cases it isn't. You'll probably want to detect your own objects: items in your production line, specific goods you sell in your store and more. For these cases, cloud services won't do the job and you'll need to train your own custom object detection model.
We can help you build a custom object detection model in minimum time using Tensorflow and popular pre-trained models.
What will we need? A labeled dataset. That's hundreds or even better if it's thousands of pictures (or videos) of what you expect to detect. You can provide the dataset or we can help you build it.
Deploy it to your own hardware (can be completely offline), you could even run this in any cloud provider or even a Raspberry Pi if you need it.
It is easy, it's a time consuming task but it is usually not hard to do.
To label a dataset you just need to draw a bounding box around the object that you want to detect. We can provide tools for that, don't worry.
We will for sure need your help if you are working with medical images or other fields that need expert knowledge to identify the objects.
Bounding boxes not enough? Don't worry, we can do pixel by pixel segmentation. That's called Image Segmentation and we can help you with that as well.