Detect an Object from long distance using ARKit and CoreML

Question

How can we detect an object from almost any distance using CoreML or ARKit. Currently I have tried ARKIT 3d object scanning and detection which works from the near distance to the object, But as we get far away from the object, it's unable to detect that object. Similarly using CoreML, I have trained a model to detect an object, which was only detecting the object from the near distance. How to achieve object detection from far distances in which actual object obviously gets too small.

Answer 1

It seems to me that scanning a far-away object has no practical application.

Firstly, you need to scan your object from all available sides in order to recognize that object in the future, not just from the only one PoV.

Secondly, there's sometimes environmental haze (smog, rain, etc) what desaturates such long-distance object and makes it hardly recognizable.

Lastly, there's a shortage of details for far-away object's surface, in other words a "texture" is quite poor.

Practical solution for this issue is simple: just use an image recognition technique for pre-taken photo in ARKit and CoreML. Also scanned data is a little bit heavier for processing than .mlmodel, isn't it?