Copy-ready prompt
A hyper-realistic screenshot of a macOS desktop showcases the workspace of a machine learning engineer at night. The image is taken from a frontal view, with a dark blue macOS menu bar at the top and the Dock visible at the bottom. Two main application windows are displayed side-by-side on the desktop. On the left is a dark-themed Visual Studio Code window occupying about two-thirds of the screen. The VS Code project, named "VISIONCLASSIFIER" in the file explorer sidebar, contains a realistic Python ML folder tree with 11 visible top-level or expanded items: .venv, data, raw, processed, images, notebooks, src, utils, config.yaml, requirements.txt, and README.md. Within the notebooks folder, two visible files are displayed: 01_data_exploration.ipynb and 02_model_training.ipynb. The src folder displays the actual ML code structure, including dataset.py, transforms.py, models, resnet.py, train, engine.py, trainer.py, and utils.py. Four tabs are open in the editor area: trainer.py, engine.py, resnet.py, and config.yaml, with trainer.py currently active. Clear and reliable Python training code for the ResNet image classification pipeline is displayed, including the Trainer class, train(self) and train_epoch(self, epoch: int) -> Dict[str, float] methods, referencing self.cfg.training.epochs, train_metrics, val_metrics, scheduler.step, save_checkpoint, self.model.train(), batch["image"], batch["label"], optimizer.zero_grad, criterion, loss.backward, optimizer.step, and accuracy(outputs, targets, topk=(1,))[0]. The code should be clear and have a natural screen feel, with line numbers displayed between lines 24 and 52. The VS Code window opens the integrated terminal's TERMINAL tab at the bottom, displaying the actual training logs for four epochs: Epoch 12/50, Epoch 13/50, Epoch 14/50, and Epoch 15/50. Each line contains training and validation data for Loss, Acc@1, and Acc@5, with the last line indicating that a new best checkpoint has been saved. The values should reflect a successful training process, with Top-1 accuracy between 0.88 and 0.91, and Top-5 accuracy between 0.97 and 0.98. The bottom includes the standard VS Code status bar, displaying Python environment details. On the right is a dark-themed web browser window displaying a local dashboard on localhost:8000, titled "VisionClassifier | Dashboard," with the application title "VisionClassifier" and the subtitle "Image Classification Model." The dashboard comprises three stacked sections. The first section, "Model Overview," includes four metric cards: Top-1 Accuracy 91.23%, Top-5 Accuracy 98.30%, Total Parameters 23.51M, and Model ResNet-50. The second section, "Recent Training," displays a dark line graph of accuracy over 50 epochs, featuring two colored curves labeled Train (Top-1) and Val (Top-1), which rise rapidly and stabilize around 90%. The third section, "Confusion Matrix," displays a 10x10 heatmap with bright diagonal lines and axes labeled True and Predicted. Utilizing subtle reflections, clear typography, realistic UI spacing, and lifelike screen halo, the macOS top menu bar displays commonly used menus such as Code, File, Edit, Selection, View, Go, Run, Terminal, Window, and Help on the left, and system icons on the right, with the time displayed as Tue May 13 9:41 AM. The Dock should contain multiple recognizable application icons, giving an overall realistic and uncluttered feel. Overall style: hyper-realistic screenshot, professional developer workstation, refined dark mode interface, unstylized, without illustration-like elements, indistinguishable from a real screen screenshot.
Prompt breakdown
A hyper-realistic screenshot of a macOS desktop showcases the workspace of a machine learning engineer at night.
The image is taken from a frontal view, with a dark blue macOS menu bar at the top and the Dock visible at the bottom.
Two main application windows are displayed side-by-side on the desktop.
On the left is a dark-themed Visual Studio Code window occupying about two-thirds of the screen.









