AI Auto Photo Cropping Process for similar photos

I am looking for a freelancer who can develop an AI Auto Photo Cropping Process. The main purpose of this process is to save time in manually cropping photos. I am attaching sample images for what I am going to be auto cropping. I have thousands of photos like these and I would like to crop them to a square with just the athlete and a little bit of padding, if available. The actual process of automatically cropping with some padding is simple, of course. The more complicated thing is automatically selecting the subject so the area that is left behind is the correct area. I have tried tools like Adobe Photoshop Select Subject (cloud version) and it works sometimes but fails at much too high of a rate and it doesn't let me input any parameters in to let the program know more about what to look for. I also cannot train it to learn and get better. If it did work, I could use the "Image Processor" function in Photoshop and/or a script like is sold on Mars Premedia. But I need to look elsewhere until it improves. I am looking for someone to help me make this sort of auto cropping a reality. I am willing and able to help train the model to tell it where the subject is. The only parameters needed for this, outside of being able to select the subject (in this case, the subject is simply the athlete head to toe along with the entire barbell, end to end --- nothing more, nothing less), is the amount of padding around the subject and the aspect ratio. In this case, the padding should be an "up to" value. ie, if there is 2% padding on all sides within the specified aspect ratio (let's say 1:1), crop to the subject, plus 2% padding. If the 2% padding doesn't exist on let's say, the bottom, then 2% padding on the three sides with the extra space with no extra space on the bottom. Hopefully this makes sense. This is important: I need the model to be able to select the subject with a 99%+ accuracy. I understand it will not be able to do this on day 1. But it needs to be something that can be trained to work. It only needs to work on the style of photos that I am attaching (ie, similar angles, similar sport, similar quality, not way out of focus, etc). If the accuracy level can only reach 90% or similar, then there's no point in doing this. I am okay with using a paid service like Amazon Rekognition or Google Cloud API or some other service (as long as the price is reasonable) but it must be much more accurate than Adobe Select Subject Cloud. And you'll have to explain to me how to make this all work because I am a photographer, not a developer. I am attaching some sample photos. These have been greatly downscaled from the original image. I am just attaching them for some examples of what I am looking to auto crop. There is only one image here per competition but I have many thousands of images from each event. I can provide more images to whoever ends up working on this project with me. Ideal Skills and Experience: - Proficiency in Machine Learning. - Experience in developing algorithms for image processing and object detection. - Ability to handle a large volume of photos. - Cannot be a service where the photos need to be uploaded to a website or similar -- I want this to happen client-side. The freelancer will be responsible for developing an AI model that can automatically detect and crop similar photos, thus improving the efficiency of the cropping process. The model should be able to handle a high volume of photos and deliver accurate results. If you read this far, please start your proposal with the phrase 2024 -- that way I know you're not an auto bidding bot who will waste both of our time. If you have the necessary skills and experience in Machine Learning and image processing, and are capable of handling a large volume of photos, please submit your proposal.