ry seeds were harvested and stored until all genotypes were available to image at the same time. On a sheet of black paper 20-30 dry seeds were uniformly distributed and imaged using Zeiss SteREO Discovery V.x fluorescence stereomicroscope using a 10x objective. Bright-field images were collected for Col-0, cmt3-11 and miR823 CRISPR knockout line. All seed size analysis were performed using Fiji software [ref fiji]. Images were converted from RGB to 8-bit and intra-class variance between images minimized by using Otsu's auto threshold clustering algorithm [ref otsu method]. These images were used as a substrate to calculate seed area using pixels as metric. For rCMT3 and gCMT3 transgenic lines segregating (T2) seeds were used to acquire both bright-field and RFP fluorescence images. Images from transgenic lines were treated as mutants except fluorescence and transmitted light channels were merged and segmented to quantify fluorescence intensity to identify RFP positive and negative candidates in segregating lines. All Fiji macros used for batch analysis of images and final seed size results are deposited at https://github.com/Papareddy/2021/tree/main/seed_size_analysis