A foreign object detection method based on the deep generative model is proposed to accurately detect the foreign objects on the coal mine belt conveyor. First, a conventional variational auto-encoder (VAE) is used to reconstruct the image, and the presence of foreign objects in the image is detected according to the reconstruction error between the original image and the reconstructed image. Considering that the reconstructed image generated by the VAE is usually fuzzy, a generative adversarial network (GAN) is introduced to evaluate the original image and the reconstructed image for a clearer image and higher foreign object detection accuracy. Finally, the VAE is combined with the GAN to design a deep learning algorithm suitable for belt foreign object detection. The experimental results show that compared with the baseline method the proposed method has a better effect on every evaluation indexes.