-
Notifications
You must be signed in to change notification settings - Fork 13
2. MEYE Models
Leonardo Lupori edited this page May 20, 2021
·
1 revision
MEYE tool is preloaded with the following CNN models, that come with different characteristics in terms of performance and minimum requirements.
| Model | Train/Test Set | mDice | FPS (Web*) | FPS (Keras**) | FLOPS | # Params |
|---|---|---|---|---|---|---|
| mini-UNet | Human/Mouse Eyes | 84.0% | 23.2 | 45.2 | 0.2G | 0.03M |
| DeepLabv3+/ResNet-50 | ImageNet + finetune on Human/Mouse Eyes | 80.1% | <1 | 28.7 | 14.1G | 26.8M |
| DeepLabv3+/Lite-MobileNet-V3-Small | ImageNet + finetune on Human/Mouse Eyes | 69.0% | 18.8 | 34.8 | 0.3G | 1.1M |
* Dell Laptop - Intel(R) Core(TM) i7-9750H CPU @ 2.60GHz - Intel UHD Graphics 630 GPU - TensorFlow.js Backend: WebGL - Browser: Microsoft Edge 90.0.818.56
** Ubuntu PC - Intel(R) Core(TM) i9-9900K CPU @ 3.60GHz - GeForce RTX 2080 Ti GPU - Python 3.6.9 + TensorFlow 2.4.1