AlexNet 1.2

By for July 3, 2018

Report Abuse
AlexNet is a convolutional neural network for classification.
AlexNet is the name of a convolutional neural network for classification, which competed in the ImageNet Large Scale Visual Recognition Challenge in 2012. **Differences:** - not training with the relighting data-augmentation; initializing - non-zero biases to 0.1 instead of 1 (found necessary for training, as initialization to 1 gave flat loss). **Object categories for classification** [Class ID][3] **Paper** [ImageNet Classification with Deep Convolutional Neural Networks][1] **Dataset** [ILSVRC2012][2] Source Caffe BVLC AlexNet ==> Caffe2 AlexNet ==> ONNX AlexNet **Model input and output** **Input** data_0: float[1, 3, 224, 224] **Output** softmaxout_1: float[1, 1000] **Sample test data** random generated sample test data: - test_data_0.npz - test_data_1.npz - test_data_2.npz - test_data_set_0 - test_data_set_1 - test_data_set_2 **Results/accuracy on test set** The bundled model is the iteration 360,000 snapshot. The best validation performance during training was iteration 358,000 with validation accuracy 57.258% and loss 1.83948. This model obtains a top-1 accuracy 57.1% and a top-5 accuracy 80.2% on the validation set, using just the center crop. (Using the average of 10 crops, (4 + 1 center) * 2 mirror, should obtain a bit higher accuracy.) **Model size: 244 MB** **License** [BSD-3][4] [1]: [2]: [3]: [4]: