Creative Technology · Computer Vision
AI Image Classification & Object Detection
Use the same image and model, then vary blur from 0px to 50px to observe how prediction confidence changes.
Choose an image
Start with a tested example, or try a photo of your own.
Start here · Example images
Four images are drawn at random from the example set.
PNG, JPEG, or WebP works best.
Use 2–20 short descriptions, separated by commas or new lines.
Run once at 0px, then again at 50px with the same image and model to compare scores.
Top-k controls how many high-scoring classes or detections are shown.
Choose an example or upload your own image.
AI result
What did the model find?
Choose an image first. It will appear here before inference.
| Rank | Label | Score |
|---|---|---|
| Run a model to see details. | ||
Blur experiment
Measure how image degradation changes model confidence.
Keep the input image and model fixed. Run at blur 0px for the baseline, then increase blur and run again. Each successful run remains in the comparison log, allowing the top label and score to be compared across blur levels.
Why blur? Image classifiers and object detectors rely on edges, textures, shapes, and fine visual details. Blur progressively removes high-frequency detail while preserving the overall scene, making it a controlled way to test model robustness. Comparing 0px with stronger blur shows whether confidence decreases gradually, whether the top label changes, and which models remain reliable when image quality is reduced.
Comparison log
Keep one image fixed and compare model decisions.
Each successful run is added below. Record the selected blur value when discussing the results.
| Run | Image | Model | Blur | Top result | Score | Time |
|---|---|---|---|---|---|---|
| Run the same image at 0px and 50px blur to start a comparison. | ||||||