The study investigated bast fibres sources: linseed flax (Linum usitatisimum L.) (cultivar-sorrel) and canola (Brassica napus L.) (cultivar- 45H29). Microstructure was examined using scanning electron microscopy (SEM) and x-ray micro-computed tomography (XMT) while retting protocol effects were assessed through gravimetric analysis. Chemical composition prediction and plant material classification were carried out using spectroscopic data (near and mid-infrared). The SEM images revealed elementary fibres of varying sizes for the canola cultivars 45H29 and L241C and flax cultivar sorrel with a width of 20±6.63 µm (n=26), 30±4.97 µm (n=8), and 29±8.12 µm (n=133), respectively. The XMT scans provided solids volume of the outer layer, excluding air space for cultivars 45H29 and sorrel, which was 3.73% and 17.2%, respectively, with n=200 images per sample. Additionally, controlled aerobic retting was carried out on both plant materials using Bacillus subtilis, subsp. spizizenii (ATCC® 6633™) as an inoculant. The retting treatments included processed (P), processed without autoclaving (P-A), processed without salts (P-S), processed without aerobic retting (P-R), and processed with inoculation (P+I). For the whole portion of flax, the treatment P+I (n=12) resulted in a significant reduction (12.5 g/100g) in neutral detergent fibre (NDFp) compared to the unprocessed (UP, n=6). Conversely, the P-S (n=4) treatment significantly increased NDFp, with a 7.49 g/100g increase compared to the P (n=16) treatment. The processed treatments, especially P and P+I, also significantly decreased D-galacturonic acid (DGAp) compared to UP. For canola, reductions in NDFp and DGAp were observed in the whole portion. The P-S treatment (n=4) showed a significant decrease in NDFp (11.2 g/100g), while the P+I (n=12) in DGAp (1.14 g/100g) when compared with the UP samples. Furthermore, the prediction of chemical composition and classification of the processed samples using peak integration as features from near and mid-infrared spectroscopy were successful. The partial least square regression models based on 32 peak integrations from near-infrared (NIR) spectra demonstrated coefficients of determination (R2) of 0.886, 0.775, 0.91, 0.847 for predicting D-galacturonic acid, hemicellulose, cellulose, and lignin, respectively. Similarly, the partial least squares discriminant analysis models for plant material classification had accuracy above 93% in classifying fibre versus shive portions and canola versus flax samples.