what is cv-variable-preview fr?

CV Variable Preview is a VS Code extension for people debugging Python code that works with images, arrays, or tables of data. Normally, if you are stepping through code and want to see what an image array actually looks like, you have to write extra lines of code to display it. This extension lets you just look at the variable directly inside the debugger, no display code needed. While paused at a breakpoint, you can hover over a variable name in your file to see a small thumbnail along with its shape and data type. Right-clicking a variable in the Variables or Watch panel opens a full preview panel where you can zoom in up to 16 times, pan around, and hover over individual pixels to see their values. It also shows a live histogram of pixel values broken down by color channel. You can compare multiple variables side by side by pinning them to a grid, which is useful when you want to check how an image changes at different steps of your code. There is also a Live mode that automatically refreshes the preview every time your debugger pauses, so you can watch values update as you step through the code. The extension understands several common data types used in image and data work: numpy arrays, PIL images, PyTorch tensors, TensorFlow eager tensors, pandas DataFrames and Series shown as scrollable tables, and lists or batches of images shown as a grid. To use it, you need VS Code 1.90 or later, a Python debug session paused at a breakpoint, and either Pillow or OpenCV installed in the Python environment you are debugging. Support for PyTorch, TensorFlow, or pandas types only kicks in if those libraries are installed too. Installation is done from a downloadable .vsix file, and after installing you need to fully quit and relaunch VS Code rather than just reloading the window. The author notes a few limitations: temporary preview image files are not automatically deleted, batches are capped at 64 items, DataFrame previews are capped at 200 rows, and only eager TensorFlow tensors are supported, not symbolic graph tensors.