Figure 1. Signaller receiving single line token from locomotive crew (Photo: Robert Baughan, 2016) A railway station scene with a steam train, staff on the tracks, a signal box, and a gift shop sign. The outdoor railway station scene shows a steam train beside a platform. Steam rises around t... More about this image found in Signaller receiving single line token from locomotive crew (Photo: Robert B...

Figure 2. Team of volunteers carrying out track renewal works (Photo: Robert Baughan, 2016) A railway worksite photo shows 4 workers on tracks with tools, sleepers, ballast, and a maintenance vehicle nearby. The railway worksite photo shows 4 workers wearing high-visibility clothing on and bes... More about this image found in Team of volunteers carrying out track renewal works (Photo: Robert Baughan,...

Figure 3. Footplate crew at work driving a steam locomotive and managing the boiler pressure (Photo: Robert Baughan, 2019) A steam train carries coal and 3 people, with steam rising from the engine as it moves along tracks through a grassy cutting. More about this image found in Footplate crew at work driving a steam locomotive and managing the boiler p...

Figure 4. Infrastructure upgrade works involving the safe and effective coordination of staff, plant, and machinery (Photo: Robert Baughan, 2018) A railway construction site with workers in high-visibility clothing, laid tracks, concrete sleepers, ballast, equipment, and an excavator beside a w... More about this image found in Infrastructure upgrade works involving the safe and effective coordination ...

Figure 5. Locomotive crew and station staff carrying out their duties to safely operate the train service (Photo: Robert Baughan, 2016) A steam locomotive stands beside a station platform with a crew member on the front, passengers nearby, carriages behind, and trees along the track. More about this image found in Locomotive crew and station staff carrying out their duties to safely opera...

Figure 6. ‘Driver’s eye’ view of the track and signals ahead (Photo: Robert Baughan, 2015) A view from a steam locomotive cab shows the boiler, metal pipes, rain-marked window, track ahead, and open fields beside the railway. More about this image found in ‘Driver’s eye’ view of the track and signals ahead (Photo: Robert Baughan, ...

Figure 7. Heritage signalling for controlling safe operation of trains (Photo: Robert Baughan, 2019) A train passes through a wooded cutting beside a tall railway signal with ladder, raised signal arms, tracks, and grassy embankments. More about this image found in Heritage signalling for controlling safe operation of trains (Photo: Robert...

Figure 1. Three-stage data pre-processing pipeline with quality control feedback loops A workflow from historical photographs, contemporary imagery and archive documents to a preprocessed dataset with quality control feedback. The workflow diagram presents three input sources on the left label... More about this image found in Three-stage data pre-processing pipeline with quality control feedback loop...

Figure 2. Integrated CV-GAN architecture for historic district reconstruction A workflow diagram traces historic imagery through feature extraction, feature fusion, a Style G A N 2 generator, and output reconstruction. The workflow diagram presents a left-to-right process. It starts with input... More about this image found in Integrated CV-GAN architecture for historic district reconstruction A wo...

Figure 3. Dataset composition analysis A two-panel chart compares architectural style counts and a documentation timeline with wartime gaps across years. Panel A is titled Architectural style distribution. It plots number of images by style. Victorian has 3,412 images, 29.0 per cent. Neoclassi... More about this image found in Dataset composition analysis A two-panel chart compares architectural st...

Figure 4. Feature preservation and style consistency analysis A paired bar chart compares architectural element detection accuracy and period-specific style preservation across generative models. The architectural element detection graph uses detection accuracy in per cent on the y-axis, rangi... More about this image found in Feature preservation and style consistency analysis A paired bar chart c...

Figure 5. Perceptual quality assessment (user study n = 45) A grouped bar chart presents mean opinion scores from 1 to 5 for five quality criteria across four generative models. The vertical axis is labelled mean opinion score: 1 to 5, with a visible range from 2.5 to 4.5. The horizontal axis lists overall quality, historical accuracy, detail preservation, environmental context, and material fidelity. The legend lists C V hyphen G A N, Style G A N 2, Pix 2 Pix, and Cycle G A N. For overall quality, C V hyphen G A N is about 3.8, Style G A N 2 is about 3.7, Pix 2 Pix is about 3.3, and Cycle G A N is about 3.3. For historical accuracy, C V hyphen G A N is about 3.4, Style G A N 2 is about 3.4, Pix 2 Pix is about 3.1, and Cycle G A N is about 3.1. For detail preservation, C V hyphen G A N is about 3.6, Style G A N 2 is about 3.5, Pix 2 Pix is about 3.2, and Cycle G A N is about 3.2. For environmental context, C V hyphen G A N is about 3.9, Style G A N 2 is about 3.7, Pix 2 Pix is about 3.5, and Cycle G A N is about 3.5. For material fidelity, C V hyphen G A N is about 3.2, Style G A N 2 is about 3.3, Pix 2 Pix is about 3.1, and Cycle G A N is about 3.2. More about this image found in Perceptual quality assessment (user study n = 45) A gro...

Figure 6. Comparative performance analysis A paired chart presents trade-off analysis for four models using metric scores, processing time, and reconstruction accuracy. The performance trade-off analysis is a radar chart with five metrics: P S N R, S S I M, F I D superscript minus 1, accuracy,... More about this image found in Comparative performance analysis A paired chart presents trade-off analy...

Figure 7. Reconstruction performance plotted against damage severity with baseline comparison A grouped bar chart gives reconstruction success rates for four models across minimal, moderate, severe, and critical building damage levels. The vertical axis is labelled reconstruction success rate ... More about this image found in Reconstruction performance plotted against damage severity with baseline co...

Figure 8. Temporal and style-specific performance evaluation A two-part chart presents F I D score over training epochs and style-specific accuracy percentages for four models. The training convergence analysis uses training epochs on the horizontal axis, ranging from 0 to 300. The vertical ax... More about this image found in Temporal and style-specific performance evaluation A two-part chart pres...

Figure 9. Representative reconstruction examples across application scenarios A comparison panel presents degraded inputs, C V hyphen G A N reconstructions, and historical references for four architectural damage cases. The panel has 3 column headers: degraded input, C V hyphen G A N reconstru... More about this image found in Representative reconstruction examples across application scenarios A co...

Figure 10. Computational performance scaling analysis: (a) training time scaling; (b) memory requirements; (c) multi-GPU scaling; (d) inference time by complexity; (e) batch processing trade-offs; and (f) computational distribution A six-panel chart compares training time, G P U memory, speedup... More about this image found in Computational performance scaling analysis: (a) training time scaling; (b) ...