ResNet is a family of deep convolutional neural networks that introduced residual (skip) connections to enable stable training of very deep architectures with strong representational capacity.

Original paper: Deep Residual Learning for Image Recognition, He et al., 2015

ResNet-50

ResNet-50 is a commonly used 50-layer variant that offers a strong balance between accuracy and computational cost and is widely adopted as a baseline and as a backbone feature extractor for tasks such as object detection, segmentation, and re-identification.

Model Configuration:

• Reference implementation: ResNet50_v1.5
• Original Weight: ResNet50_Weights.IMAGENET1K_V2
• Resolution: 3x224x224
• Support Cooper version:
  • Cooper SDK: [2.5.3]
  • Cooper Foundry: [2.2]

Model	Device	compression	Model Link
Resnet50	N1-655	Amba_optimized	Model_Link
Resnet50	N1-655	Activation_fp16	Model_Link
Resnet50	CV7	Amba_optimized	Model_Link
Resnet50	CV7	Activation_fp16	Model_Link
Resnet50	CV72	Amba_optimized	Model_Link
Resnet50	CV72	Activation_fp16	Model_Link
Resnet50	CV75	Amba_optimized	Model_Link
Resnet50	CV75	Activation_fp16	Model_Link