Every architecture available in ConvLens, plotted by accuracy, speed, and parameter count. Use this page to pick the right model for your task — then open the viewer to see what it learned.
Lowest CPU latency — ideal for real-time demos and mobile.
Strongest accuracy-per-millisecond tradeoff.
Highest ImageNet top-1 accuracy in the registry.
Highest accuracy per million parameters.
Bubble size represents parameter count. Models toward the upper-left offer the best tradeoff.
| Model | Top-1 acc | CPU latency | Params (M) |
|---|---|---|---|
| ResNet-50 | 80.9% | 80 ms | 25.6 |
| ResNet-18 | 69.8% | 30 ms | 11.7 |
| Model | Top-1 acc | CPU latency | Params (M) |
|---|---|---|---|
| MobileNet V3 Large | 75.3% | 30 ms | 5.5 |
| MobileNet V2 | 72.2% | 25 ms | 3.5 |
| MobileNet V3 Small | 67.7% | 15 ms | 2.5 |
| Model | Top-1 acc | CPU latency | Params (M) |
|---|---|---|---|
| EfficientNet-B3 | 82.0% | 110 ms | 12.2 |
| EfficientNet-B2 | 80.6% | 70 ms | 9.1 |
| EfficientNet-B0 | 77.7% | 40 ms | 5.3 |
| Model | Top-1 acc | CPU latency | Params (M) |
|---|---|---|---|
| DenseNet-121 | 74.4% | 70 ms | 8.0 |
| Model | Top-1 acc | CPU latency | Params (M) |
|---|---|---|---|
| ConvNeXt-Tiny | 82.5% | 100 ms | 28.6 |
| Model | Top-1 acc | CPU latency | Params (M) |
|---|---|---|---|
| ShuffleNet V2 | 69.4% | 20 ms | 2.3 |
Top-1 ImageNet accuracy figures are reproduced from the torchvision model zoo using the default pretrained weights. CPU inference latency is an approximate measurement taken on the ConvLens backend container (single-thread, 224×224 input, warm cache) and will vary with hardware. Parameter counts are reported by torchvision and rounded to one decimal.