Abstract: The digital image of industrial radiographic films is the basis for realizing the informationization and intelligence of non-destructive testing and defect assessment processes. Since the film digitization process is affected by film blackness quality and human operations, it is easy to produce a large number of unqualified digital images. How to quickly and quantitatively evaluate the quality of digital images has become an urgent bottleneck problem that needs to be solved. To solve the above problems, an end-to-end rapid evaluation technique for film digital image quality that integrates deep learning theory was proposed. Firstly, a digital image dataset based on the binary classification principle was built using a large amount of actually existing industrial ray digital image data. Then, a digital image binary classfication evaluation method based on improved VGGNet was constructed. The method was illustrated and verified by the radiographic scanning data of a pipeline girth weld of a certain enterprise. The results show that the proposed method can achieve end-to-end image quality evaluation with an accuracy of 90.0%-95.7%.