In order to identify the corrosion properties of three kinds of austenitic stainless steel used in solar power plant, experiments were conducted to test the weight loss after dynamic immersing in molten salt, the corrosion thickness rates per year was obtained, and the surface morphology by SEM and component of corrosive product by EDS were analyzed. The test result showed that the corrosion resistance of 316L and 347H are outstanding, much better than 304. The corrosion resistance order of the three material is 304 <316L< 347H. 304 and 316L can be considered as the candidate materials of low-temperature molten salt storage tank, and 347H can be used as the materials of high-temperature molten salt storage tank and molten salt heat exchanger.
Multi-label image classification is a foundational topic in various domains. Multimodal learning approaches have recently achieved outstanding results in image representation and single-label image classification. For instance, Contrastive Language-Image Pretraining (CLIP) demonstrates impressive image-text representation learning abilities and is robust to natural distribution shifts. This success inspires us to leverage multimodal learning for multi-label classification tasks, and benefit from contrastively learnt pretrained models. We propose the Multimodal Multi-label Image Classification (MuMIC) framework, which utilizes a hardness-aware tempered sigmoid based Binary Cross Entropy loss function, thus enables the optimization on multi-label objectives and transfer learning on CLIP. MuMIC is capable of providing high classification performance, handling real-world noisy data, supporting zero-shot predictions, and producing domain-specific image embeddings. In this study, a total of 120 image classes are defined, and more than 140K positive annotations are collected on approximately 60K Booking.com images. The final MuMIC model is deployed on Booking.com Content Intelligence Platform, and it outperforms other state-of-the-art models with 85.6% GAP@10 and 83.8% GAP on all 120 classes, as well as a 90.1% macro mAP score across 32 majority classes. We summarize the modeling choices which are extensively tested through ablation studies. To the best of our knowledge, we are the first to adapt contrastively learnt multimodal pretraining for real-world multi-label image classification problems, and the innovation can be transferred to other domains.
The aim was to explore the application value of computed tomography (CT) images based on optimized reconstruction algorithm (ORA) in knee osteoarthritis (KOA) and to analyze the rehabilitation effect of exercise therapy combined with celecoxib capsules on KOA. On the basis of iterative reconstruction algorithm, sparse data and compressed sensing (CS) algorithm were introduced in this study for the image reconstruction and were compared with the reconstruction effects of other algorithms. 105 patients with KOA were selected as the subjects investigated, who were treated in the hospital from December 2018 to January 2020. They were divided randomly into group A (exercise therapy), group B (celecoxib capsule therapy), and group C (exercise therapy combined with celecoxib capsules), with 35 patients in each group. All the patients were scanned by CT. Visual analogue score (VAS) method was adopted to compare the pain symptoms of patients in the three groups before and after treatment. The Lysholm knee function score was employed to evaluate the knee function of the patient before and after treatment, and finally the difference in the total effective rate of treatment was compared. The results showed that the reconstruction algorithm in this study had a higher degree of coincidence with the original image, and its root mean square error (RMSE), peak signal-to-noise ratio (PSNR), and structural similarity (SSIM) were greatly higher than those of filtered back-projection (FBP) algorithm. There was no huge difference between Lysholm score and VAS in patients of the three groups before treatment ( P > 0.05 ). Besides, Lysholm score in all dimensions and its total score after treatment were higher than those of patients in the three groups before treatment ( P < 0.05 ), and those of patients in group C increased extremely in contrast to group A ( P < 0.05 ). VAS after treatment were all lower in patients of the three groups than those before treatment ( P < 0.05 ), and VAS of patients in group C were sharply lower than those of group A ( P < 0.05 ), but the total effective rate of patients in group C was dramatically higher than the rate of group A ( P < 0.05 ). It indicated that an algorithm with short scanning time and good image quality was established based on iterative reconstruction algorithm. Exercise therapy combined with celecoxib capsules could alleviate the pain symptoms and improve the function of knee joint, which was more advantageous than the simple rehabilitation comprehensive therapy. Therefore, it had higher clinical value for the patients with KOA.