AI-Based Object Detection Project

We would like to hire someone who can build us a high-performance “Onscreen Object Detector” in C++. The AI is tasked to search the current screen for an object and report its position. Preferably the project should come as a single DLL with exported functions as following (example):

struct Vector3 {float x,y,z};
Vector3 GetClosestObjectToCenter() // returns a single position of an object that is the closest to the center
int GetObjectCount() // returns the count of objects found on the screen
void FindObjectOnCurrentScreen() // checks the current screen/display for objects


The project should consist of 2 components:

Component 1 – Learning Module
In the learning module we can mark a set of images with the desired object to be found.
Requirements:
- No pre-defined requirements as long as it works.
- A proper designed workflow to create new models is desired though.

Component 2 – Live Search Module
This module will use the trained model from component 1 in order to find an object if existing on the screen.

Requirements:
- Must be fast (30FPS+)
- Binary/C++ Component only so it can be integrated with 3rd Party software.

If you apply for this job please include a roughly estimate on
- on how many days you require for a PoC/Prototype
- on how many days you require to complete the project
- what software you would use (e.g. Tensorflow/NVidia etc.)
- very rough estimate of total costs.
- include your hourly rate for further developing the project after completion.


The project is complete when:
- Component 1 has been delivered and we are able to train new models based on marking custom images.
- Component 2 is able to use the newly trained model and return positions.

We are aware that the performance of component 2 heavily depends on the used hardware. As for reference a Windows system with a NVidia 1080 should be able to produce at least 30 FPS.