---
library_name: pytorch
---

![EfficientNetV2_logo](resource/EfficientNetV2.png)

EfficientNetV2 proposes improved scaling rules and training-aware architectural optimizations, including fused-MBConv blocks, to achieve faster training and better accuracy–efficiency trade-offs than prior EfficientNet models.

Original paper: [EfficientNetV2: Smaller Models and Faster Training](https://arxiv.org/abs/2104.00298)

# EfficientNetV2-S

This model uses the EfficientNetV2-S variant, a compact configuration that balances accuracy, inference latency, and training speed. It is well suited for production image classification and as a backbone in vision pipelines where fast convergence and efficient deployment are important.

Model Configuration:
- Reference implementation: [torchvision.models.efficientnet_v2_s](https://pytorch.org/vision/stable/models/generated/torchvision.models.efficientnet_v2_s.html)
- Original Weight: [EfficientNet_V2_S_Weights.IMAGENET1K_V1](https://download.pytorch.org/models/efficientnet_v2_s-dd5fe13b.pth)
- Resolution: 3x384x384
- Support Cooper version:
    - Cooper SDK: [2.5.2]
    - Cooper Foundry: [2.2]

| Model | Device | Model Link |
| :-----: | :-----: | :-----: |
| EfficientNetV2 | N1-655 | [Model_Link](https://huggingface.co/Ambarella/EfficientNetV2/blob/main/n1-655_efficientnet_v2_s.bin) |
| EfficientNetV2 | CV72 | [Model_Link](https://huggingface.co/Ambarella/EfficientNetV2/blob/main/cv72_efficientnet_v2_s.bin) |
| EfficientNetV2 | CV75 | [Model_Link](https://huggingface.co/Ambarella/EfficientNetV2/blob/main/cv75_efficientnet_v2_s.bin) |