Polyethylene fibers (PEF) are useful for manufacturing high tenacity concrete with strain hardening and multi-cracking behavior. However, there are few quantitative studies on the complex cracking characteristics of PEF reinforced cement-based materials. In this paper, the external crack fractal dimension (CFD) of PEF-reinforced alkali activated mortar (PEFRAAM) after flexural failure was studied. The cracks on the PEFRAAM could be accurately extracted by the deep learning network. The crack segmentation results were post-processed to obtain the CFD. Experimental results indicated that the CFD of PEFRAAM was increased with the growth of the volume fraction of PEF. The PEF length was more useful than water-binder ratio (WBR) of mortar to mitigate the CFD value increase. Consequently, the maximum CFD was reached in the alkali activated mortar with a WBR of 0.48 and a PEF length of 12 mm. The CFD of PEFRAAM was improved with the growth of PEF reinforcing index in a quadratic function. The critical reinforcing index and density reinforcing index were 150 and 400, respectively. Moreover, the CFD of PEFRAAMs had an obvious positive relation with the flexural strength and the toughness in an exponential function. The excellent correlation between CFD and the flexural toughness before peak load was discovered, resulting from the good correlation between the pre-peak deflection hardening behavior and multi-crack cracking behavior. The full flexural toughness and flexural toughness before peak load could be divided into three regions, whose division thresholds were approximately 1.0 and 1.2. The variation trend of CFD on the flank could reflect the overall cracking pattern of beam, reducing the difficulty to obtain bottom crack information in experiment and engineering.