Best way to train yolov5 on a custom dataset

Question

I have a dataset with about 100 images that look like this.

My goal is get yolov5 to detect buildings in similar images. In order to do this I would like yolov5 to get to close to 1 in precision on the training dataset The images are of size 10000 pixels x 10000 pixels I have tried following this tutorial: https://towardsdatascience.com/how-to-train-a-custom-object-detection-model-with-yolo-v5-917e9ce13208

But I notice that there are some differences. First, I have images with 10000 pixels that means I should run

!python train.py --img  10000 -batch 10 --epochs 5000 --data '/content/yolov5/data.yaml' --cfg ./models/custom_yolov5s.yaml --weights '' --name yolov5s_results  --cache

But I get an error

WARNING: --img-size 10000 must be multiple of max stride 32, updating to 10016enter code here

So my question is:

1. Is yolov5 appropriate for my task?
2. How should I run the !python train.py command? What's are a recommended batch and epoch size?
3. Is there anything else that I should change in the actual yolov5 backbone to make it more appropriate for my task?