GoogLeNet for Image Classification
- TensorFlow implementation of Going Deeper with Convolutions (CVPR'15).
- This repository contains the examples of natural image classification using pre-trained model as well as training a Inception network from scratch on CIFAR-10 dataset (93.64% accuracy on testing set). The pre-trained model on CIFAR-10 can be download from here.
- Architecture of GoogLeNet from the paper:
Requirements
- Python 3.3+
- TensorFlow 1.9+
- Numpy
- Scipy
Implementation Details
- The GoogLeNet model is defined in
src/nets/googlenet.py. - Inception module is defined in
src/models/inception_module.py. - An example of image classification using pre-trained model is in
examples/inception_pretrained.py. - An example of train a network from scratch on CIFAR-10 is in
examples/inception_cifar.py.
For testing the pre-trained model
- Images are rescaled so that the smallest side equals 224 before fed into the model. This is not the same as the original paper which is an ensemble of 7 similar models using 144 224x224 crops per image for testing. So the performance will not be as good as the original paper.
For training from scratch on CIFAR-10
- All the LRN layers are removed from the convolutional layers.
- Batch normalization and ReLU activation are used in all the convolutional layers including the layers in Inception structure except the output layer.
- Two auxiliary classifiers are used as mentioned in the paper, though 512 instead of 1024 hidden units are used in the two fully connected layers to reduce the computation. However, I found the results are almost the same on CIFAR-10 with and without auxiliary classifiers.
- Since the 32 x 32 images are down-sampled to 1 x 1 before fed into
inception_5a, this makes the multi-scale structure of inception layers less useful and harm the performance (around 80% accuracy). To make full use of the multi-scale structures, the stride of the first convolutional layer is reduced to 1 and the first two max pooling layers are removed. The the feature map (32 x 32 x channels) will have almost the same size as described in table 1 (28 x 28 x channel) in the paper before fed intoinception_3a. I have also tried only reduce the stride or only remove one max pooling layer. But I found the current setting provides the best performance on the testing set. - During training, dropout with keep probability 0.4 is applied to two fully connected layers and weight decay with 5e-4 is used as well.
- The network is trained through Adam optimizer. Batch size is 128. The initial learning rate is 1e-3, decays to 1e-4 after 30 epochs, and finally decays to 1e-5 after 50 epochs.
- Each color channel of the input images are subtracted by the mean value computed from the training set.
Usage
ImageNet Classification
Preparation
- Download the pre-trained parameters here. This is original from here.
- Setup path in
examples/inception_pretrained.py:PRETRINED_PATHis the path for pre-trained model.DATA_PATHis the path to put testing images.
Run
Go to examples/ and put test image in folder DATA_PATH, then run the script:
python inception_pretrained.py --im_name PART_OF_IMAGE_NAME
--im_nameis the option for image names you want to test. If the testing images are allpngfiles, this can bepng. The default setting is.jpg.- The output will be the top-5 class labels and probabilities.
Train the network on CIFAR-10
Preparation
- Download CIFAR-10 dataset from here
- Setup path in
examples/inception_cifar.py:DATA_PATHis the path to put CIFAR-10.SAVE_PATHis the path to save or load summary file and trained model.
Train the model
Go to examples/ and run the script:
python inception_cifar.py --train \
--lr LEARNING_RATE \
--bsize BATCH_SIZE \
--keep_prob KEEP_PROB_OF_DROPOUT \
--maxepoch MAX_TRAINING_EPOCH
- Summary and model will be saved in
SAVE_PATH. One pre-trained model on CIFAR-10 can be downloaded from here.
Evaluate the model
Go to examples/ and put the pre-trained model in SAVE_PATH. Then run the script:
python inception_cifar.py --eval \
--load PRE_TRAINED_MODEL_ID
- The pre-trained ID is epoch ID shown in the save modeled file name. The default value is
99, which indicates the one I uploaded. - The output will be the accuracy of training and testing set.
Results
Image classification using pre-trained model
- Top five predictions are shown. The probabilities are shown keeping two decimal places. Note that the pre-trained model are trained on ImageNet.
- Result of VGG19 for the same images can be found here.
The pre-processing of images for both experiments are the same.
Data Source