resnet

ResNet

Deeper neural networks are more difficult to train. Residual learning framework ease the training of networks that are substantially deeper. The research explicitly reformulate the layers as learning residual functions with reference to the layer inputs, instead of learning unreferenced functions. It also provide comprehensive empirical evidence showing that these residual networks are easier to optimize, and can gain accuracy from considerably increased depth. On the ImageNet dataset the residual nets were evaluated with a depth of up to 152 layers — 8× deeper than VGG nets but still having lower complexity. The models perform image classification - they take images as input and classifies the major object in the image into a set of pre-defined classes. They are trained ImageNet dataset which contains images from 1000 classes. ResNet models provide very high accuracies with affordable model sizes. They are ideal for cases when high accuracy of classification is required.

Model

The model below are ResNet v1 and v2. ResNet models consists of residual blocks and came up to counter the effect of deteriorating accuracies with more layers due to network not learning the initial layers. ResNet v2 uses pre-activation function whereas ResNet v1 uses post-activation for the residual blocks. The models below have 18, 34, 50, 101 and 152 layers for with ResNetv1 and ResNetv2 architecture.

• Version 1:

Model	ONNX Model	Model archives	Top-1 accuracy (%)	Top-5 accuracy (%)
ResNet-18	44.7 MB	44.7 MB	69.93	89.29
ResNet-34	83.3 MB	83.4 MB	73.73	91.40
ResNet-50	97.8 MB	98.0 MB	74.93	92.38
ResNet-101	170.6 MB	170.9 MB	76.48	93.20
ResNet-152	230.6 MB	230.9 MB	77.11	93.61

• Version 2:

Model	ONNX Model	Model archives	Top-1 accuracy (%)	Top-5 accuracy (%)
ResNet-18	44.6 MB	44.7 MB	69.70	89.49
ResNet-34	83.2 MB	83.3 MB	73.36	91.43
ResNet-50	97.7 MB	97.8 MB	75.81	92.82
ResNet-101	170.4 MB	170.6 MB	77.42	93.61
ResNet-152	230.3 MB	230.6 MB	78.20	94.21

Inference

We used MXNet as framework with gluon APIs to perform inference. View the notebook imagenet_inference to understand how to use above models for doing inference. Make sure to specify the appropriate model name in the notebook.

Input

All pre-trained models expect input images normalized in the same way, i.e. mini-batches of 3-channel RGB images of shape (N x 3 x H x W), where N is the batch size, and H and W are expected to be at least 224. The inference was done using jpeg image.

Preprocessing

The images have to be loaded in to a range of [0, 1] and then normalized using mean = [0.485, 0.456, 0.406] and std = [0.229, 0.224, 0.225]. The transformation should preferrably happen at preprocessing.

import mxnet
from mxnet.gluon.data.vision import transforms
def preprocess(img):   
    '''
    
    Preprocessing required on the images for inference with mxnet gluon
    The function takes path to an image and returns processed tensor
    ''''''
    transform_fn = transforms.Compose([
    transforms.Resize(224),
    transforms.CenterCrop(224),
    transforms.ToTensor(),
    transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
    ])
    img = transform_fn(img)
    img = img.expand_dims(axis=0) # batchify
    
    return img
    

Output

The model outputs image scores for each of the 1000 classes of ImageNet.

Postprocessing

The post-processing involves calculating the softmax probablility scores for each classes and sorting them to report the most probable classes.

import mxnet as mx
def postprocess(scores): 
    '''
    Postprocessing with mxnet gluon
    The function takes scores generated by network and returns the class IDs in decreasing order
    of probability
    ''''''
    prob = mx.ndarray.softmax(scores).asnumpy()
    prob = np.squeeze(prob)
    a = np.argsort(prob)[::-1]
    return a
    

Inference with Model Server

To learn how to use model archives with Model Server, try out the Model Server QuickStart to get Model Server installed and tested. If you already have installed the server, you can use the commands below to start serving this model.

• Start Server:

mxnet-model-server --models resnet18v1=https://s3.amazonaws.com/onnx-model-zoo/resnet/resnet18v1/resnet18v1.model

• Run Prediction:

curl -O https://s3.amazonaws.com/model-server/inputs/kitten.jpg
curl -X POST http://127.0.0.1:8080/resnet18v1/predict -F "[email protected]"

For inference requests with all above ResNet models, Model Server expects the image to be passed in the data variable, which is the input layer's name in the model. In the previous example this was [email protected].

Dataset

Dataset used for train and validation: ImageNet (ILSVRC2012). Check imagenet_prep for guidelines on preparing the dataset.

Validation accuracy

The accuracies obtained by the models on the validation set are mentioned above. The validation was done using center cropping of images unlike the paper using ten-cropping. Even with center crop, the accuracies are within 1-2% of accuracy reported by the paper.

Training

We used MXNet as framework with gluon APIs to perform training. View the training notebook to understand details for parameters and network for each of the above variants of ResNet.

Validation

We used MXNet as framework with gluon APIs to perform validation. Use the notebook imagenet_validation to verify the accuracy of the model on the validation set. Make sure to specify the appropriate model name in the notebook.

References

ResNet-v1 Deep residual learning for image recognition He, Kaiming, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. In Proceedings of the IEEE conference on computer vision and pattern recognition, pp. 770-778. 2016.

ResNet-v2 Identity mappings in deep residual networks He, Kaiming, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. In European Conference on Computer Vision, pp. 630-645. Springer, Cham, 2016.

Contributors

• ankkhedia (Amazon AI)
• abhinavs95 (Amazon AI)

Acknowledgments

MXNet, Gluon model zoo, GluonCV, MMS

Keywords

CNN, ResNet, ONNX, ImageNet, Computer Vision