resnet50

torchvision.models.quantization.resnet50(*, weights: Optional[Union[ResNet50_QuantizedWeights, ResNet50_Weights]] = None, progress: bool = True, quantize: bool = False, **kwargs: Any) → QuantizableResNet

ResNet-50 model from Deep Residual Learning for Image Recognition

Note

Note that quantize = True returns a quantized model with 8 bit weights. Quantized models only support inference and run on CPUs. GPU inference is not yet supported.

Parameters:

• weights (ResNet50_QuantizedWeights or ResNet50_Weights, optional) – The pretrained weights for the model. See ResNet50_QuantizedWeights below for more details, and possible values. By default, no pre-trained weights are used.

• progress (bool, optional) – If True, displays a progress bar of the download to stderr. Default is True.

• quantize (bool, optional) – If True, return a quantized version of the model. Default is False.

• **kwargs – parameters passed to the torchvision.models.quantization.QuantizableResNet base class. Please refer to the source code for more details about this class.

class torchvision.models.quantization.ResNet50_QuantizedWeights(value)

The model builder above accepts the following values as the weights parameter. ResNet50_QuantizedWeights.DEFAULT is equivalent to ResNet50_QuantizedWeights.IMAGENET1K_FBGEMM_V2. You can also use strings, e.g. weights='DEFAULT' or weights='IMAGENET1K_FBGEMM_V1'.

ResNet50_QuantizedWeights.IMAGENET1K_FBGEMM_V1:

These weights were produced by doing Post Training Quantization (eager mode) on top of the unquantized weights listed below.

acc@1 (on ImageNet-1K)	75.92
acc@5 (on ImageNet-1K)	92.814
min_size	height=1, width=1
categories	tench, goldfish, great white shark, … (997 omitted)
backend	fbgemm
recipe	link
num_params	25557032
unquantized	ResNet50_Weights.IMAGENET1K_V1
GIPS	4.09
File size	24.8 MB

The inference transforms are available at ResNet50_QuantizedWeights.IMAGENET1K_FBGEMM_V1.transforms and perform the following preprocessing operations: Accepts PIL.Image, batched (B, C, H, W) and single (C, H, W) image torch.Tensor objects. The images are resized to resize_size=[256] using interpolation=InterpolationMode.BILINEAR, followed by a central crop of crop_size=[224]. Finally the values are first rescaled to [0.0, 1.0] and then normalized using mean=[0.485, 0.456, 0.406] and std=[0.229, 0.224, 0.225].

ResNet50_QuantizedWeights.IMAGENET1K_FBGEMM_V2:

These weights were produced by doing Post Training Quantization (eager mode) on top of the unquantized weights listed below. Also available as ResNet50_QuantizedWeights.DEFAULT.

acc@1 (on ImageNet-1K)	80.282
acc@5 (on ImageNet-1K)	94.976
min_size	height=1, width=1
categories	tench, goldfish, great white shark, … (997 omitted)
backend	fbgemm
recipe	link
num_params	25557032
unquantized	ResNet50_Weights.IMAGENET1K_V2
GIPS	4.09
File size	25.0 MB

The inference transforms are available at ResNet50_QuantizedWeights.IMAGENET1K_FBGEMM_V2.transforms and perform the following preprocessing operations: Accepts PIL.Image, batched (B, C, H, W) and single (C, H, W) image torch.Tensor objects. The images are resized to resize_size=[232] using interpolation=InterpolationMode.BILINEAR, followed by a central crop of crop_size=[224]. Finally the values are first rescaled to [0.0, 1.0] and then normalized using mean=[0.485, 0.456, 0.406] and std=[0.229, 0.224, 0.225].

class torchvision.models.ResNet50_Weights(value)

The model builder above accepts the following values as the weights parameter. ResNet50_Weights.DEFAULT is equivalent to ResNet50_Weights.IMAGENET1K_V2. You can also use strings, e.g. weights='DEFAULT' or weights='IMAGENET1K_V1'.

ResNet50_Weights.IMAGENET1K_V1:

These weights reproduce closely the results of the paper using a simple training recipe.

acc@1 (on ImageNet-1K)	76.13
acc@5 (on ImageNet-1K)	92.862
min_size	height=1, width=1
categories	tench, goldfish, great white shark, … (997 omitted)
num_params	25557032
recipe	link
GFLOPS	4.09
File size	97.8 MB

The inference transforms are available at ResNet50_Weights.IMAGENET1K_V1.transforms and perform the following preprocessing operations: Accepts PIL.Image, batched (B, C, H, W) and single (C, H, W) image torch.Tensor objects. The images are resized to resize_size=[256] using interpolation=InterpolationMode.BILINEAR, followed by a central crop of crop_size=[224]. Finally the values are first rescaled to [0.0, 1.0] and then normalized using mean=[0.485, 0.456, 0.406] and std=[0.229, 0.224, 0.225].

ResNet50_Weights.IMAGENET1K_V2:

These weights improve upon the results of the original paper by using TorchVision’s new training recipe. Also available as ResNet50_Weights.DEFAULT.

acc@1 (on ImageNet-1K)	80.858
acc@5 (on ImageNet-1K)	95.434
min_size	height=1, width=1
categories	tench, goldfish, great white shark, … (997 omitted)
num_params	25557032
recipe	link
GFLOPS	4.09
File size	97.8 MB

The inference transforms are available at ResNet50_Weights.IMAGENET1K_V2.transforms and perform the following preprocessing operations: Accepts PIL.Image, batched (B, C, H, W) and single (C, H, W) image torch.Tensor objects. The images are resized to resize_size=[232] using interpolation=InterpolationMode.BILINEAR, followed by a central crop of crop_size=[224]. Finally the values are first rescaled to [0.0, 1.0] and then normalized using mean=[0.485, 0.456, 0.406] and std=[0.229, 0.224, 0.225].