Either look at the position of the flanking size standard bands and make best estimate...almost always good enough since we often know the band size expected or use the size standard to plot 1/distance travelled against band size which should be quite linear and read off your band distance against that standard line. Not commonly done but it works. Expensive transilluminator software packages do the calculation for you but are probably only 5% better than judging it by eye
If I remember well. There is a linear relation between Log(DNA lenght) and dinstance (in mm). You can meassure the distance traveled by each band of your DNA ladder, then convert fragment size to Log scale and plot the results. You should see a negative correlation between the two. Solve the line equation for Log(DNA length) and use it to input the distance of your unknown samples. Then transform the result ( Log(DNA length)) to decimal scale.