Learning hash codes for efficient content reuse detection