Can text-independent writer identification be done without multi-sentence training datasets for each writer?

I have a folder with 100k images, each of them with a different handwritten sentence. All of the images have sentences of different lengths. They range from about 30 to 80 English characters. The file names start at 1.png and go up to 100k.png. That's it, as far as input data. 95% of the sentences are written by different writers. 5% are written by the same writers. Some writers might have written 2 sentences, while others 300+.