Railways speedily transport many people and goods nationwide, so railway accidents can pose immense damage. However, the infrastructure of railways is so complex that its maintenance is challenging and expensive. Therefore, using artificial intelligence for railway safety has attracted many researchers. This paper examines artificial intelligence applications for railway safety, mainly focusing on deep learning approaches. This paper first introduces deep learning methods widely used for railway safety. Then, we investigated and classified earlier studies into four representative application areas: (1) railway infrastructure (catenary, surface, components, and geometry), (2) train body and bogie (door, wheel, suspension, bearing, etc.), (3) operation (railway detection, railroad trespassing, wind risk, train running safety, etc.), and (4) station (air quality control, accident prevention, etc.). We present fundamental problems and popular approaches for each application area. Finally, based on the literature reviews, we discuss the opportunities and challenges of artificial intelligence for railway safety.

Pose embeddings provide a practical way to improve viewpoint invariance and occlusion robustness in 2D pose analysis by transforming sparse landmark coordinates into representations that preserve underlying 3D pose semantics. Existing approaches often rely on contrastive or variational objectives, which require careful pair construction, large batches, or additional training constraints. In this paper, we propose masked pose embedding (MPE), a transformer-based autoencoder trained with a single masked-reconstruction objective tailored to pose landmarks. MPE combines two key components: (1) view-angle canonicalization, in which randomly rotated 3D landmarks are projected to 2D and decoded toward the original canonical pose, and (2) landmark-level masking with masked-only supervision to improve robustness to partial visibility. We evaluate MPE in two complementary settings. First, on SARA and Human3.6M, we analyze whether distances in the learned embedding space correlate with ground-truth 3D pose distances under both controlled perturbations and under natural, unperturbed data. Second, on Yoga-82, we assess real image-based pose retrieval using YOLOv8x-pose landmarks under multiple synthetic occlusion patterns. The results show that MPE more faithfully preserves 3D pose semantics than recent single-frame and transformer-based representation baselines and achieves stronger retrieval accuracy among the compared methods under both full and partial landmark observations. These findings suggest that MPE provides an effective and training-stable solution for view- and occlusion-robust 2D pose embedding, while remaining highly parameter-efficient compared with recent single-frame and temporal representation baselines.

Camera-based affective sensing degrades sharply when the face is partially occluded by hands, masks, eyeglasses or headwear. Existing occlusion-robust facial expression recognition (FER) methods either discard the information carried by occluded regions through attention re-weighting, hallucinate it back at the pixel level, or restore it in a purely visual latent space; none of them can consult an external semantic description of what the hidden region ought to look like. We propose TeCoRe, a text-conditioned completion-and-recognition framework that restores occluded patch tokens directly inside the joint vision–language embedding space of a frozen CLIP encoder. A lightweight visibility gate estimates a per-patch occlusion probability; a two-layer transformer decoder reconstructs the occluded tokens using class-level textual emotion embeddings, produced by the frozen CLIP text encoder, as both cross-attention memory and an alignment target; and a gated blend passes clean tokens through unchanged so that completion cannot corrupt visible evidence. TeCoRe is trained as supervised FER using the ground-truth emotion labels for classification and text alignment, together with an auxiliary self-supervised reconstruction objective. Because the occlusion masks are synthesised during training, the pre-occlusion CLIP tokens of the same image provide the reconstruction target without requiring additional manual annotation. The CLIP backbone is completely frozen—no low-rank adapters, no landmarks, no segmentation priors—and only about 27 M head and decoder parameters are trained. On the standard RAN occlusion and pose subsets, TeCoRe reaches 91.42% on Occlusion-RAF-DB, 84.79% on Occlusion-FERPlus, and 62.42% on Occlusion-AffectNet. It achieves the highest accuracy in our comparison on Occlusion-RAF-DB and on the RAF-DB and FERPlus pose subsets, while remaining competitive but not leading on the FERPlus and AffectNet occlusion subsets. The ablation and control studies are reported as three-seed means so that the small differences between variants are not read off a single initialisation. A backbone-generality control shows positive mean gains from the proposed modules across all three evaluated frozen CLIP configurations, with the largest gain observed in the main ViT-L/14@336 setting. These results indicate that performing occluded-token completion in a joint vision–language space, with class-level text embeddings providing shared conditioning to the completion process, is an effective and parameter-efficient route to occlusion-robust FER.