Tensorflow object detection project / Consultancy

I'm currently working in a personal project where I want to: Use an IP camera (rtsp capable) and connect to it using a raspberry pi 3 (raspberry pi camera v2 will be used for testing) with motion ([login to view URL]) and detect if there are movements in the frame, if so save the picture to a folder, and have a python script checking if there is a picture in the folder, then process it with tensorflow object detection and get the detected object (if any) and it's probability and then delete the picture. This is intended to detect when cars move around the place where I live (small town), it's for the security since we have been facing some robberies in the neighborhood, the final objective is to take note of the car's license plate, I have the ALPR (automatic license plate recognition) part covered using a web service that works perfectly, but I need to upload a picture of a car showing the plate license, since I can't upload all the pictures when I detect movements because the web service charge me per every picture processed then I decided to make this personal project, since not all cars pictures work I want to first check if the picture has a car showing a license plate before upload it to the web service, so in order to do that I want to train a model with several cars photos (of cars showing the license plate, I have like 1000) so the algorithm can not only detect if there is a car in the picture, but if it's showing a license plate which means it's a valid picture for me and I can upload it to the web service. I was thinking on use tensorflow ssdlite_mobilenet_v2_coco as base model, but since I won't be processing a live feed (all the frames) and only processing specific images that won't be there all the time, maybe I can use a model with better accuracy like ssd_inception_v2_coco or even faster_rcnn_resnet101_coco? (don't know if the last one works in the raspberry pi, you are suppose to guide here). Also I know raspberry pi supports tensorflow-lite which considerably reduces the load and improve the processing speed, so I also want to learn how to properly setup it. There are many many tutorials available and different ways to do what I need to do so I was thinking I can get the help from a professional to the following tasks: A. Teach me / guide me (via email, no video calls required, like you send me the guide I follow it and if I have trouble then I can come back to you and you can help me ): 1. How to properly setup tensorflow in windows 10 and both tensorflow & tensorflow-lite in the raspberry pi 3 2. How to train a model in my Windows computer (I have a MSI GTX 1650 VENTUS XS OC 4GB graphic card that I can use to improve the process) including the setup of tensorflow-gpu and CUDA and all required to use the GPU for training. 3. How to convert a model to tensorflow-lite the right way (some people say you can see a big decrease in accuracy if it's not done properly) B. Create a template Python script that I can pass a picture to it (as an argument) and it can print out the object detect by tensorflow / tensorflow lite (if any) and it's probability. I know c programming and some python too, so I'm not a newbie but I don't have the time to dive into all the tutorials hehehe, hope an experienced professional can help me. Please only bid if you are experienced, can prove it and can accomplish 100% the required tasks.