DECORE · Compressed vs. Full VGG-16
One reinforcement-learning agent per channel learns which to drop. All three models run entirely in your browser — no server, no GPU.
| Metric | Full VGG-16 | DECORE-pruned | DECORE + INT8 |
|---|---|---|---|
| Parameters | 14.72 M | 5.43 M | 5.43 M |
| Model size | 58.9 MB | 21.8 MB | 5.5 MB |
| Top-1 accuracy | 93.48% | 93.76% | 93.68% |
| vs. full size | 1× | 2.7× smaller | ~11× smaller |
or upload your own image
Examples (CIFAR-10 test set)
Full VGG-16
14.7M paramsDECORE-pruned
5.4M paramsDECORE + INT8 final deliverable
5.5 MB · INT8💰 What would this save you?
Same accuracy, ~2× faster end-to-end, 63% smaller. A model's real-world worth is load time + inference time combined — load is what a cold-starting serverless replica or a first-time visitor pays, inference is what every request pays. Enter your current spend to estimate the impact on your bill.
Load — · Inference — · Total —
Load + inference times are measured live in your browser (from page load and from running an image above). Savings assume compute cost scales with combined load+inference time at constant request volume — a directional estimate, not a quote.