Example of training and evaluation of ResNet50 on Imagenet

This example trains a ResNet50 model on the ImageNet2012 dataset.

Getting data

You have to obtain the Imagenet dataset to train the model.

Internally this code uses TFDS which will show download instructions on the first run. Run python examples/image_classification/imagenet_resnet50_train.py and you will see download instructions, similar to the following:

AssertionError: Manual directory /home/${USER}/tensorflow_datasets/downloads/manual does not exist or is empty. Create it and download/extract dataset artifacts in there. Additional instructions: manual_dir should contain two files: ILSVRC2012_img_train.tar and
ILSVRC2012_img_val.tar.

You have to download data from http://www.image-net.org/download-images and then put it into the directory mentioned in the message. On the next run, run imagenet_resnet50_train.py which will process the data and rearrange it inside the data directory which might take a while. Subsequent runs will re-use the already downloaded data.

You can override TFDS data directory by providing the --tfds_data_dir flag. This might be useful if you don't have enough disk space in the default location or already have a copy of Imagenet data somewhere else.

Training the model

Use the following command to train:

python examples/classify/img/imagenet/imagenet_train.py \
    --model_dir="${HOME}/experiments/resnet50"

Some additional useful flags include the following:

• --train_device_batch_size controls per-device training batch size. You may need to adjust it if you don't have enough GPU memory.
• --eval_device_batch_size controls per-device evaluation batch size. You may need to adjust it if you don't have enough GPU memory.
• --eval_every_n_steps controls the number of training steps between evaluation and checkpointing.
• --tfds_data_dir overrides the directory where TFDS looks for datasets.