Supplementary Video 2. Fast, versatile and quantitative annotation of complex images BatesKathleen JiangShen ChaudharyShivesh Jackson-HolmesEmily JueMelinda L McCaskeyErin GoldmanDaniel I LuHang 2019 <div><p>Supplementary Video 2. Fast, versatile and quantitative annotation of complex images </p></div><div><p>Representative example reconstruction of worm behavior. Segmented images from raw data are on left, with purely algorithmic reconstruction in center, and consensus annotation-based reconstruction on right (only for complex or ambiguous frames). Although algorithmic reconstruction is visually more accurate in many frames, annotation-based reconstruction produces accurate results even in frames that are inaccurately segmented or in cases where the algorithm performs poorly. Movie is at half speed from original 30 fps. </p></div>