Evaluation of mining productivity, including the determination of the geometric dimensions of rock objects in an open pit, is one of the most critical tasks in the mining industry. The problem of rock fragmentation is usually solved using computer vision methods such as instance segmentation or semantic segmentation. Today, deep learning neural networks are used to solve such problems for digital images. Neural networks require a lot of computing power to process high-resolution digital images and large datasets. To address this issue, in literature, lightweight architectural neural networks are proposed, as well as parallel computing using CPU, GPU, and specialized accelerators. The review discusses the latest advances in the field of deep learning neural networks for solving computer vision problems in relation to rock fragmentation and aspects of improving the performance of neural network implementations on various parallel architectures.