Accurate estimation of material weight transported by belt conveyors is crucial for monitoring operational efficiency in various industries. Traditional weight measurement methods using physical sensors are often unreliable due to high installation and maintenance costs, as well as susceptibility to noise, errors, and faults. This paper proposes a deep learning-based method to address these issues by estimating the weight of material transported by a belt conveyor using real-time current measurements from the conveyor’s engines. The model is trained on time-series data from multiple conveyors equipped with weight measurement systems. The primary goal is to develop a model that generalizes well across various operating conditions, providing a reliable alternative to physical weight measurement systems. The convolutional neural network model incorporates residual and attention blocks and is trained on a dataset of 5,650 work shifts. The results demonstrate that the model achieves an R2 score of 0.8952 and an RMSE of 0.0745 on the validation dataset, and its estimations are statistically more accurate than physical weight sensor measurements, improving the R2 by 0.4963 and reducing the MAPE by 12.86%. The contribution of the paper is to present a data-driven methodology employing a convolutional neural network to effectively estimate material weight on belt conveyors, addressing the limitations of existing weight measurement techniques.