gitrepostats

#!/bin/sh
set -e
get_all_objects () {
        git rev-list --all --objects |  git cat-file --batch-check='%(objectname) %(objecttype) %(objectsize) %(rest)'
}
get_all_objects | awk '
  BEGIN {
    size_limit = 10*1024*1024;
  }
  $2 == "blob" {
    c += 1; s += $3
    if ($3 > size_limit) {
      tops += $3;
      topc += 1;
      bf[$4] = $3;
    }
  }
  END {
    big_file_percentage = 100.0*tops/s;
    s = s/(1024.0*1024.0);
    tops = tops/(1024.0*1024.0);
    printf "file count:     %d\nfile size:     %.1f MB\n", c, s
    printf "big file count: %d\nbig file size: %.1f MB (%.1f%%)\n", topc, tops, big_file_percentage
    printf "big file definition: size > %.1f MB\n", size_limit/(1024.0*1024.0)
    for (name in bf)
       printf "%-10d %s\n", bf[name]/1024, name | "sort -k1 -n -r"
    close("sort -k1 -n -r")
  }'

Find out which files occupy most of the disk space in the git repository.
Unlike du this explores the whole history (not just the current working copy)